ops.py 69.4 KB
Newer Older
Q
qingqing01 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
#   Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import paddle
import paddle.nn.functional as F
import paddle.nn as nn
from paddle import ParamAttr
from paddle.regularizer import L2Decay

from paddle.fluid.framework import Variable, in_dygraph_mode
from paddle.fluid import core
from paddle.fluid.layer_helper import LayerHelper
from paddle.fluid.dygraph import layers
from paddle.fluid.data_feeder import check_variable_and_dtype, check_type, check_dtype, convert_dtype
import math
import six
import numpy as np
from functools import reduce

__all__ = [
    'roi_pool',
    'roi_align',
    'prior_box',
    'generate_proposals',
    'iou_similarity',
    'box_coder',
    'yolo_box',
    'multiclass_nms',
    'distribute_fpn_proposals',
    'collect_fpn_proposals',
    'matrix_nms',
    'batch_norm',
W
wangxinxin08 已提交
44
    'mish',
Q
qingqing01 已提交
45 46 47
]


W
wangxinxin08 已提交
48 49 50 51
def mish(x):
    return x * paddle.tanh(F.softplus(x))


52 53 54
def batch_norm(ch,
               norm_type='bn',
               norm_decay=0.,
55
               freeze_norm=False,
56 57
               initializer=None,
               data_format='NCHW'):
Q
qingqing01 已提交
58 59 60 61 62
    if norm_type == 'sync_bn':
        batch_norm = nn.SyncBatchNorm
    else:
        batch_norm = nn.BatchNorm2D

63 64 65 66 67 68 69 70 71 72 73 74
    norm_lr = 0. if freeze_norm else 1.
    weight_attr = ParamAttr(
        initializer=initializer,
        learning_rate=norm_lr,
        regularizer=L2Decay(norm_decay),
        trainable=False if freeze_norm else True)
    bias_attr = ParamAttr(
        learning_rate=norm_lr,
        regularizer=L2Decay(norm_decay),
        trainable=False if freeze_norm else True)

    norm_layer = batch_norm(
Q
qingqing01 已提交
75
        ch,
76 77
        weight_attr=weight_attr,
        bias_attr=bias_attr,
78
        data_format=data_format)
Q
qingqing01 已提交
79

80 81 82 83 84 85 86
    norm_params = norm_layer.parameters()
    if freeze_norm:
        for param in norm_params:
            param.stop_gradient = True

    return norm_layer

Q
qingqing01 已提交
87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194

@paddle.jit.not_to_static
def roi_pool(input,
             rois,
             output_size,
             spatial_scale=1.0,
             rois_num=None,
             name=None):
    """

    This operator implements the roi_pooling layer.
    Region of interest pooling (also known as RoI pooling) is to perform max pooling on inputs of nonuniform sizes to obtain fixed-size feature maps (e.g. 7*7).

    The operator has three steps:

        1. Dividing each region proposal into equal-sized sections with output_size(h, w);
        2. Finding the largest value in each section;
        3. Copying these max values to the output buffer.

    For more information, please refer to https://stackoverflow.com/questions/43430056/what-is-roi-layer-in-fast-rcnn

    Args:
        input (Tensor): Input feature, 4D-Tensor with the shape of [N,C,H,W], 
            where N is the batch size, C is the input channel, H is Height, W is weight. 
            The data type is float32 or float64.
        rois (Tensor): ROIs (Regions of Interest) to pool over. 
            2D-Tensor or 2D-LoDTensor with the shape of [num_rois,4], the lod level is 1. 
            Given as [[x1, y1, x2, y2], ...], (x1, y1) is the top left coordinates, 
            and (x2, y2) is the bottom right coordinates.
        output_size (int or tuple[int, int]): The pooled output size(h, w), data type is int32. If int, h and w are both equal to output_size.
        spatial_scale (float, optional): Multiplicative spatial scale factor to translate ROI coords from their input scale to the scale used when pooling. Default: 1.0
        rois_num (Tensor): The number of RoIs in each image. Default: None
        name(str, optional): For detailed information, please refer
            to :ref:`api_guide_Name`. Usually name is no need to set and
            None by default.


    Returns:
        Tensor: The pooled feature, 4D-Tensor with the shape of [num_rois, C, output_size[0], output_size[1]].


    Examples:

    ..  code-block:: python

        import paddle
        from ppdet.modeling import ops
        paddle.enable_static()

        x = paddle.static.data(
                name='data', shape=[None, 256, 32, 32], dtype='float32')
        rois = paddle.static.data(
                name='rois', shape=[None, 4], dtype='float32')
        rois_num = paddle.static.data(name='rois_num', shape=[None], dtype='int32')

        pool_out = ops.roi_pool(
                input=x,
                rois=rois,
                output_size=(1, 1),
                spatial_scale=1.0,
                rois_num=rois_num)
    """
    check_type(output_size, 'output_size', (int, tuple), 'roi_pool')
    if isinstance(output_size, int):
        output_size = (output_size, output_size)

    pooled_height, pooled_width = output_size
    if in_dygraph_mode():
        assert rois_num is not None, "rois_num should not be None in dygraph mode."
        pool_out, argmaxes = core.ops.roi_pool(
            input, rois, rois_num, "pooled_height", pooled_height,
            "pooled_width", pooled_width, "spatial_scale", spatial_scale)
        return pool_out, argmaxes

    else:
        check_variable_and_dtype(input, 'input', ['float32'], 'roi_pool')
        check_variable_and_dtype(rois, 'rois', ['float32'], 'roi_pool')
        helper = LayerHelper('roi_pool', **locals())
        dtype = helper.input_dtype()
        pool_out = helper.create_variable_for_type_inference(dtype)
        argmaxes = helper.create_variable_for_type_inference(dtype='int32')

        inputs = {
            "X": input,
            "ROIs": rois,
        }
        if rois_num is not None:
            inputs['RoisNum'] = rois_num
        helper.append_op(
            type="roi_pool",
            inputs=inputs,
            outputs={"Out": pool_out,
                     "Argmax": argmaxes},
            attrs={
                "pooled_height": pooled_height,
                "pooled_width": pooled_width,
                "spatial_scale": spatial_scale
            })
        return pool_out, argmaxes


@paddle.jit.not_to_static
def roi_align(input,
              rois,
              output_size,
              spatial_scale=1.0,
              sampling_ratio=-1,
              rois_num=None,
195
              aligned=True,
Q
qingqing01 已提交
196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265
              name=None):
    """

    Region of interest align (also known as RoI align) is to perform
    bilinear interpolation on inputs of nonuniform sizes to obtain 
    fixed-size feature maps (e.g. 7*7)

    Dividing each region proposal into equal-sized sections with
    the pooled_width and pooled_height. Location remains the origin
    result.

    In each ROI bin, the value of the four regularly sampled locations 
    are computed directly through bilinear interpolation. The output is
    the mean of four locations.
    Thus avoid the misaligned problem. 

    Args:
        input (Tensor): Input feature, 4D-Tensor with the shape of [N,C,H,W], 
            where N is the batch size, C is the input channel, H is Height, W is weight. 
            The data type is float32 or float64.
        rois (Tensor): ROIs (Regions of Interest) to pool over.It should be
            a 2-D Tensor or 2-D LoDTensor of shape (num_rois, 4), the lod level is 1. 
            The data type is float32 or float64. Given as [[x1, y1, x2, y2], ...],
            (x1, y1) is the top left coordinates, and (x2, y2) is the bottom right coordinates.
        output_size (int or tuple[int, int]): The pooled output size(h, w), data type is int32. If int, h and w are both equal to output_size.
        spatial_scale (float32, optional): Multiplicative spatial scale factor to translate ROI coords 
            from their input scale to the scale used when pooling. Default: 1.0
        sampling_ratio(int32, optional): number of sampling points in the interpolation grid. 
            If <=0, then grid points are adaptive to roi_width and pooled_w, likewise for height. Default: -1
        rois_num (Tensor): The number of RoIs in each image. Default: None
        name(str, optional): For detailed information, please refer
            to :ref:`api_guide_Name`. Usually name is no need to set and
            None by default.

    Returns:
        Tensor:

        Output: The output of ROIAlignOp is a 4-D tensor with shape (num_rois, channels, pooled_h, pooled_w). The data type is float32 or float64.


    Examples:
        .. code-block:: python

            import paddle
            from ppdet.modeling import ops
            paddle.enable_static()

            x = paddle.static.data(
                name='data', shape=[None, 256, 32, 32], dtype='float32')
            rois = paddle.static.data(
                name='rois', shape=[None, 4], dtype='float32')
            rois_num = paddle.static.data(name='rois_num', shape=[None], dtype='int32')
            align_out = ops.roi_align(input=x,
                                               rois=rois,
                                               ouput_size=(7, 7),
                                               spatial_scale=0.5,
                                               sampling_ratio=-1,
                                               rois_num=rois_num)
    """
    check_type(output_size, 'output_size', (int, tuple), 'roi_align')
    if isinstance(output_size, int):
        output_size = (output_size, output_size)

    pooled_height, pooled_width = output_size

    if in_dygraph_mode():
        assert rois_num is not None, "rois_num should not be None in dygraph mode."
        align_out = core.ops.roi_align(
            input, rois, rois_num, "pooled_height", pooled_height,
            "pooled_width", pooled_width, "spatial_scale", spatial_scale,
W
wangguanzhong 已提交
266
            "sampling_ratio", sampling_ratio, "aligned", aligned)
Q
qingqing01 已提交
267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
        return align_out

    else:
        check_variable_and_dtype(input, 'input', ['float32', 'float64'],
                                 'roi_align')
        check_variable_and_dtype(rois, 'rois', ['float32', 'float64'],
                                 'roi_align')
        helper = LayerHelper('roi_align', **locals())
        dtype = helper.input_dtype()
        align_out = helper.create_variable_for_type_inference(dtype)
        inputs = {
            "X": input,
            "ROIs": rois,
        }
        if rois_num is not None:
            inputs['RoisNum'] = rois_num
        helper.append_op(
            type="roi_align",
            inputs=inputs,
            outputs={"Out": align_out},
            attrs={
                "pooled_height": pooled_height,
                "pooled_width": pooled_width,
                "spatial_scale": spatial_scale,
291
                "sampling_ratio": sampling_ratio,
W
wangguanzhong 已提交
292
                "aligned": aligned,
Q
qingqing01 已提交
293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477
            })
        return align_out


@paddle.jit.not_to_static
def iou_similarity(x, y, box_normalized=True, name=None):
    """
    Computes intersection-over-union (IOU) between two box lists.
    Box list 'X' should be a LoDTensor and 'Y' is a common Tensor,
    boxes in 'Y' are shared by all instance of the batched inputs of X.
    Given two boxes A and B, the calculation of IOU is as follows:

    $$
    IOU(A, B) = 
    \\frac{area(A\\cap B)}{area(A)+area(B)-area(A\\cap B)}
    $$

    Args:
        x (Tensor): Box list X is a 2-D Tensor with shape [N, 4] holds N 
             boxes, each box is represented as [xmin, ymin, xmax, ymax], 
             the shape of X is [N, 4]. [xmin, ymin] is the left top 
             coordinate of the box if the input is image feature map, they
             are close to the origin of the coordinate system. 
             [xmax, ymax] is the right bottom coordinate of the box.
             The data type is float32 or float64.
        y (Tensor): Box list Y holds M boxes, each box is represented as 
             [xmin, ymin, xmax, ymax], the shape of X is [N, 4]. 
             [xmin, ymin] is the left top coordinate of the box if the 
             input is image feature map, and [xmax, ymax] is the right 
             bottom coordinate of the box. The data type is float32 or float64.
        box_normalized(bool): Whether treat the priorbox as a normalized box.
            Set true by default.
        name(str, optional): For detailed information, please refer 
            to :ref:`api_guide_Name`. Usually name is no need to set and 
            None by default. 

    Returns:
        Tensor: The output of iou_similarity op, a tensor with shape [N, M] 
              representing pairwise iou scores. The data type is same with x.

    Examples:
        .. code-block:: python

            import paddle
            from ppdet.modeling import ops
            paddle.enable_static()

            x = paddle.static.data(name='x', shape=[None, 4], dtype='float32')
            y = paddle.static.data(name='y', shape=[None, 4], dtype='float32')
            iou = ops.iou_similarity(x=x, y=y)
    """

    if in_dygraph_mode():
        out = core.ops.iou_similarity(x, y, 'box_normalized', box_normalized)
        return out
    else:
        helper = LayerHelper("iou_similarity", **locals())
        out = helper.create_variable_for_type_inference(dtype=x.dtype)

        helper.append_op(
            type="iou_similarity",
            inputs={"X": x,
                    "Y": y},
            attrs={"box_normalized": box_normalized},
            outputs={"Out": out})
        return out


@paddle.jit.not_to_static
def collect_fpn_proposals(multi_rois,
                          multi_scores,
                          min_level,
                          max_level,
                          post_nms_top_n,
                          rois_num_per_level=None,
                          name=None):
    """
    
    **This OP only supports LoDTensor as input**. Concat multi-level RoIs 
    (Region of Interest) and select N RoIs with respect to multi_scores. 
    This operation performs the following steps:

    1. Choose num_level RoIs and scores as input: num_level = max_level - min_level
    2. Concat multi-level RoIs and scores
    3. Sort scores and select post_nms_top_n scores
    4. Gather RoIs by selected indices from scores
    5. Re-sort RoIs by corresponding batch_id

    Args:
        multi_rois(list): List of RoIs to collect. Element in list is 2-D 
            LoDTensor with shape [N, 4] and data type is float32 or float64, 
            N is the number of RoIs.
        multi_scores(list): List of scores of RoIs to collect. Element in list 
            is 2-D LoDTensor with shape [N, 1] and data type is float32 or
            float64, N is the number of RoIs.
        min_level(int): The lowest level of FPN layer to collect
        max_level(int): The highest level of FPN layer to collect
        post_nms_top_n(int): The number of selected RoIs
        rois_num_per_level(list, optional): The List of RoIs' numbers. 
            Each element is 1-D Tensor which contains the RoIs' number of each 
            image on each level and the shape is [B] and data type is 
            int32, B is the number of images. If it is not None then return 
            a 1-D Tensor contains the output RoIs' number of each image and 
            the shape is [B]. Default: None
        name(str, optional): For detailed information, please refer 
            to :ref:`api_guide_Name`. Usually name is no need to set and 
            None by default.

    Returns:
        Variable:

        fpn_rois(Variable): 2-D LoDTensor with shape [N, 4] and data type is 
        float32 or float64. Selected RoIs. 

        rois_num(Tensor): 1-D Tensor contains the RoIs's number of each 
        image. The shape is [B] and data type is int32. B is the number of 
        images. 

    Examples:
        .. code-block:: python
           
            import paddle
            from ppdet.modeling import ops
            paddle.enable_static()
            multi_rois = []
            multi_scores = []
            for i in range(4):
                multi_rois.append(paddle.static.data(
                    name='roi_'+str(i), shape=[None, 4], dtype='float32', lod_level=1))
            for i in range(4):
                multi_scores.append(paddle.static.data(
                    name='score_'+str(i), shape=[None, 1], dtype='float32', lod_level=1))

            fpn_rois = ops.collect_fpn_proposals(
                multi_rois=multi_rois, 
                multi_scores=multi_scores,
                min_level=2, 
                max_level=5, 
                post_nms_top_n=2000)
    """
    check_type(multi_rois, 'multi_rois', list, 'collect_fpn_proposals')
    check_type(multi_scores, 'multi_scores', list, 'collect_fpn_proposals')
    num_lvl = max_level - min_level + 1
    input_rois = multi_rois[:num_lvl]
    input_scores = multi_scores[:num_lvl]

    if in_dygraph_mode():
        assert rois_num_per_level is not None, "rois_num_per_level should not be None in dygraph mode."
        attrs = ('post_nms_topN', post_nms_top_n)
        output_rois, rois_num = core.ops.collect_fpn_proposals(
            input_rois, input_scores, rois_num_per_level, *attrs)
        return output_rois, rois_num

    else:
        helper = LayerHelper('collect_fpn_proposals', **locals())
        dtype = helper.input_dtype('multi_rois')
        check_dtype(dtype, 'multi_rois', ['float32', 'float64'],
                    'collect_fpn_proposals')
        output_rois = helper.create_variable_for_type_inference(dtype)
        output_rois.stop_gradient = True

        inputs = {
            'MultiLevelRois': input_rois,
            'MultiLevelScores': input_scores,
        }
        outputs = {'FpnRois': output_rois}
        if rois_num_per_level is not None:
            inputs['MultiLevelRoIsNum'] = rois_num_per_level
            rois_num = helper.create_variable_for_type_inference(dtype='int32')
            rois_num.stop_gradient = True
            outputs['RoisNum'] = rois_num
        helper.append_op(
            type='collect_fpn_proposals',
            inputs=inputs,
            outputs=outputs,
            attrs={'post_nms_topN': post_nms_top_n})
        return output_rois, rois_num


@paddle.jit.not_to_static
def distribute_fpn_proposals(fpn_rois,
                             min_level,
                             max_level,
                             refer_level,
                             refer_scale,
478
                             pixel_offset=False,
Q
qingqing01 已提交
479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552
                             rois_num=None,
                             name=None):
    """
    
    **This op only takes LoDTensor as input.** In Feature Pyramid Networks 
    (FPN) models, it is needed to distribute all proposals into different FPN 
    level, with respect to scale of the proposals, the referring scale and the 
    referring level. Besides, to restore the order of proposals, we return an 
    array which indicates the original index of rois in current proposals. 
    To compute FPN level for each roi, the formula is given as follows:
    
    .. math::

        roi\_scale &= \sqrt{BBoxArea(fpn\_roi)}

        level = floor(&\log(\\frac{roi\_scale}{refer\_scale}) + refer\_level)

    where BBoxArea is a function to compute the area of each roi.

    Args:

        fpn_rois(Variable): 2-D Tensor with shape [N, 4] and data type is 
            float32 or float64. The input fpn_rois.
        min_level(int32): The lowest level of FPN layer where the proposals come 
            from.
        max_level(int32): The highest level of FPN layer where the proposals
            come from.
        refer_level(int32): The referring level of FPN layer with specified scale.
        refer_scale(int32): The referring scale of FPN layer with specified level.
        rois_num(Tensor): 1-D Tensor contains the number of RoIs in each image. 
            The shape is [B] and data type is int32. B is the number of images.
            If it is not None then return a list of 1-D Tensor. Each element 
            is the output RoIs' number of each image on the corresponding level
            and the shape is [B]. None by default.
        name(str, optional): For detailed information, please refer 
            to :ref:`api_guide_Name`. Usually name is no need to set and 
            None by default. 

    Returns:
        Tuple:

        multi_rois(List) : A list of 2-D LoDTensor with shape [M, 4] 
        and data type of float32 and float64. The length is 
        max_level-min_level+1. The proposals in each FPN level.

        restore_ind(Variable): A 2-D Tensor with shape [N, 1], N is 
        the number of total rois. The data type is int32. It is
        used to restore the order of fpn_rois.

        rois_num_per_level(List): A list of 1-D Tensor and each Tensor is 
        the RoIs' number in each image on the corresponding level. The shape 
        is [B] and data type of int32. B is the number of images


    Examples:
        .. code-block:: python

            import paddle
            from ppdet.modeling import ops
            paddle.enable_static()
            fpn_rois = paddle.static.data(
                name='data', shape=[None, 4], dtype='float32', lod_level=1)
            multi_rois, restore_ind = ops.distribute_fpn_proposals(
                fpn_rois=fpn_rois,
                min_level=2,
                max_level=5,
                refer_level=4,
                refer_scale=224)
    """
    num_lvl = max_level - min_level + 1

    if in_dygraph_mode():
        assert rois_num is not None, "rois_num should not be None in dygraph mode."
        attrs = ('min_level', min_level, 'max_level', max_level, 'refer_level',
553 554
                 refer_level, 'refer_scale', refer_scale, 'pixel_offset',
                 pixel_offset)
Q
qingqing01 已提交
555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592
        multi_rois, restore_ind, rois_num_per_level = core.ops.distribute_fpn_proposals(
            fpn_rois, rois_num, num_lvl, num_lvl, *attrs)
        return multi_rois, restore_ind, rois_num_per_level

    else:
        check_variable_and_dtype(fpn_rois, 'fpn_rois', ['float32', 'float64'],
                                 'distribute_fpn_proposals')
        helper = LayerHelper('distribute_fpn_proposals', **locals())
        dtype = helper.input_dtype('fpn_rois')
        multi_rois = [
            helper.create_variable_for_type_inference(dtype)
            for i in range(num_lvl)
        ]

        restore_ind = helper.create_variable_for_type_inference(dtype='int32')

        inputs = {'FpnRois': fpn_rois}
        outputs = {
            'MultiFpnRois': multi_rois,
            'RestoreIndex': restore_ind,
        }

        if rois_num is not None:
            inputs['RoisNum'] = rois_num
            rois_num_per_level = [
                helper.create_variable_for_type_inference(dtype='int32')
                for i in range(num_lvl)
            ]
            outputs['MultiLevelRoIsNum'] = rois_num_per_level

        helper.append_op(
            type='distribute_fpn_proposals',
            inputs=inputs,
            outputs=outputs,
            attrs={
                'min_level': min_level,
                'max_level': max_level,
                'refer_level': refer_level,
593 594
                'refer_scale': refer_scale,
                'pixel_offset': pixel_offset
Q
qingqing01 已提交
595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838
            })
        return multi_rois, restore_ind, rois_num_per_level


@paddle.jit.not_to_static
def yolo_box(
        x,
        origin_shape,
        anchors,
        class_num,
        conf_thresh,
        downsample_ratio,
        clip_bbox=True,
        scale_x_y=1.,
        name=None, ):
    """

    This operator generates YOLO detection boxes from output of YOLOv3 network.
     
     The output of previous network is in shape [N, C, H, W], while H and W
     should be the same, H and W specify the grid size, each grid point predict
     given number boxes, this given number, which following will be represented as S,
     is specified by the number of anchors. In the second dimension(the channel
     dimension), C should be equal to S * (5 + class_num), class_num is the object
     category number of source dataset(such as 80 in coco dataset), so the
     second(channel) dimension, apart from 4 box location coordinates x, y, w, h,
     also includes confidence score of the box and class one-hot key of each anchor
     box.
     Assume the 4 location coordinates are :math:`t_x, t_y, t_w, t_h`, the box
     predictions should be as follows:
     $$
     b_x = \\sigma(t_x) + c_x
     $$
     $$
     b_y = \\sigma(t_y) + c_y
     $$
     $$
     b_w = p_w e^{t_w}
     $$
     $$
     b_h = p_h e^{t_h}
     $$
     in the equation above, :math:`c_x, c_y` is the left top corner of current grid
     and :math:`p_w, p_h` is specified by anchors.
     The logistic regression value of the 5th channel of each anchor prediction boxes
     represents the confidence score of each prediction box, and the logistic
     regression value of the last :attr:`class_num` channels of each anchor prediction
     boxes represents the classifcation scores. Boxes with confidence scores less than
     :attr:`conf_thresh` should be ignored, and box final scores is the product of
     confidence scores and classification scores.
     $$
     score_{pred} = score_{conf} * score_{class}
     $$

    Args:
        x (Tensor): The input tensor of YoloBox operator is a 4-D tensor with shape of [N, C, H, W].
                    The second dimension(C) stores box locations, confidence score and
                    classification one-hot keys of each anchor box. Generally, X should be the output of YOLOv3 network.
                    The data type is float32 or float64.
        origin_shape (Tensor): The image size tensor of YoloBox operator, This is a 2-D tensor with shape of [N, 2].
                    This tensor holds height and width of each input image used for resizing output box in input image
                    scale. The data type is int32.
        anchors (list|tuple): The anchor width and height, it will be parsed pair by pair.
        class_num (int): The number of classes to predict.
        conf_thresh (float): The confidence scores threshold of detection boxes. Boxes with confidence scores
                    under threshold should be ignored.
        downsample_ratio (int): The downsample ratio from network input to YoloBox operator input,
                    so 32, 16, 8 should be set for the first, second, and thrid YoloBox operators.
        clip_bbox (bool): Whether clip output bonding box in Input(ImgSize) boundary. Default true.
        scale_x_y (float): Scale the center point of decoded bounding box. Default 1.0.
        name (string): The default value is None.  Normally there is no need
                       for user to set this property.  For more information,
                       please refer to :ref:`api_guide_Name`

    Returns:
        boxes Tensor: A 3-D tensor with shape [N, M, 4], the coordinates of boxes,  N is the batch num,
                    M is output box number, and the 3rd dimension stores [xmin, ymin, xmax, ymax] coordinates of boxes.
        scores Tensor: A 3-D tensor with shape [N, M, :attr:`class_num`], the coordinates of boxes,  N is the batch num,
                    M is output box number.
                    
    Raises:
        TypeError: Attr anchors of yolo box must be list or tuple
        TypeError: Attr class_num of yolo box must be an integer
        TypeError: Attr conf_thresh of yolo box must be a float number

    Examples:

    .. code-block:: python

        import paddle
        from ppdet.modeling import ops
        
        paddle.enable_static()
        x = paddle.static.data(name='x', shape=[None, 255, 13, 13], dtype='float32')
        img_size = paddle.static.data(name='img_size',shape=[None, 2],dtype='int64')
        anchors = [10, 13, 16, 30, 33, 23]
        boxes,scores = ops.yolo_box(x=x, img_size=img_size, class_num=80, anchors=anchors,
                                        conf_thresh=0.01, downsample_ratio=32)
    """
    helper = LayerHelper('yolo_box', **locals())

    if not isinstance(anchors, list) and not isinstance(anchors, tuple):
        raise TypeError("Attr anchors of yolo_box must be list or tuple")
    if not isinstance(class_num, int):
        raise TypeError("Attr class_num of yolo_box must be an integer")
    if not isinstance(conf_thresh, float):
        raise TypeError("Attr ignore_thresh of yolo_box must be a float number")

    if in_dygraph_mode():
        attrs = ('anchors', anchors, 'class_num', class_num, 'conf_thresh',
                 conf_thresh, 'downsample_ratio', downsample_ratio, 'clip_bbox',
                 clip_bbox, 'scale_x_y', scale_x_y)
        boxes, scores = core.ops.yolo_box(x, origin_shape, *attrs)
        return boxes, scores
    else:
        boxes = helper.create_variable_for_type_inference(dtype=x.dtype)
        scores = helper.create_variable_for_type_inference(dtype=x.dtype)

        attrs = {
            "anchors": anchors,
            "class_num": class_num,
            "conf_thresh": conf_thresh,
            "downsample_ratio": downsample_ratio,
            "clip_bbox": clip_bbox,
            "scale_x_y": scale_x_y,
        }

        helper.append_op(
            type='yolo_box',
            inputs={
                "X": x,
                "ImgSize": origin_shape,
            },
            outputs={
                'Boxes': boxes,
                'Scores': scores,
            },
            attrs=attrs)
        return boxes, scores


@paddle.jit.not_to_static
def prior_box(input,
              image,
              min_sizes,
              max_sizes=None,
              aspect_ratios=[1.],
              variance=[0.1, 0.1, 0.2, 0.2],
              flip=False,
              clip=False,
              steps=[0.0, 0.0],
              offset=0.5,
              min_max_aspect_ratios_order=False,
              name=None):
    """

    This op generates prior boxes for SSD(Single Shot MultiBox Detector) algorithm.
    Each position of the input produce N prior boxes, N is determined by
    the count of min_sizes, max_sizes and aspect_ratios, The size of the
    box is in range(min_size, max_size) interval, which is generated in
    sequence according to the aspect_ratios.

    Parameters:
       input(Tensor): 4-D tensor(NCHW), the data type should be float32 or float64.
       image(Tensor): 4-D tensor(NCHW), the input image data of PriorBoxOp,
            the data type should be float32 or float64.
       min_sizes(list|tuple|float): the min sizes of generated prior boxes.
       max_sizes(list|tuple|None): the max sizes of generated prior boxes.
            Default: None.
       aspect_ratios(list|tuple|float): the aspect ratios of generated
            prior boxes. Default: [1.].
       variance(list|tuple): the variances to be encoded in prior boxes.
            Default:[0.1, 0.1, 0.2, 0.2].
       flip(bool): Whether to flip aspect ratios. Default:False.
       clip(bool): Whether to clip out-of-boundary boxes. Default: False.
       step(list|tuple): Prior boxes step across width and height, If
            step[0] equals to 0.0 or step[1] equals to 0.0, the prior boxes step across
            height or weight of the input will be automatically calculated.
            Default: [0., 0.]
       offset(float): Prior boxes center offset. Default: 0.5
       min_max_aspect_ratios_order(bool): If set True, the output prior box is
            in order of [min, max, aspect_ratios], which is consistent with
            Caffe. Please note, this order affects the weights order of
            convolution layer followed by and does not affect the final
            detection results. Default: False.
       name(str, optional): The default value is None.  Normally there is no need for 
            user to set this property. For more information, please refer to :ref:`api_guide_Name`

    Returns:
        Tuple: A tuple with two Variable (boxes, variances)

        boxes(Tensor): the output prior boxes of PriorBox.
        4-D tensor, the layout is [H, W, num_priors, 4].
        H is the height of input, W is the width of input,
        num_priors is the total box count of each position of input.

        variances(Tensor): the expanded variances of PriorBox.
        4-D tensor, the layput is [H, W, num_priors, 4].
        H is the height of input, W is the width of input
        num_priors is the total box count of each position of input

    Examples:
        .. code-block:: python

        import paddle
        from ppdet.modeling import ops

        paddle.enable_static()
        input = paddle.static.data(name="input", shape=[None,3,6,9])
        image = paddle.static.data(name="image", shape=[None,3,9,12])
        box, var = ops.prior_box(
                    input=input,
                    image=image,
                    min_sizes=[100.],
                    clip=True,
                    flip=True)
    """
    helper = LayerHelper("prior_box", **locals())
    dtype = helper.input_dtype()
    check_variable_and_dtype(
        input, 'input', ['uint8', 'int8', 'float32', 'float64'], 'prior_box')

    def _is_list_or_tuple_(data):
        return (isinstance(data, list) or isinstance(data, tuple))

    if not _is_list_or_tuple_(min_sizes):
        min_sizes = [min_sizes]
    if not _is_list_or_tuple_(aspect_ratios):
        aspect_ratios = [aspect_ratios]
    if not (_is_list_or_tuple_(steps) and len(steps) == 2):
        raise ValueError('steps should be a list or tuple ',
                         'with length 2, (step_width, step_height).')

    min_sizes = list(map(float, min_sizes))
    aspect_ratios = list(map(float, aspect_ratios))
    steps = list(map(float, steps))

    cur_max_sizes = None
    if max_sizes is not None and len(max_sizes) > 0 and max_sizes[0] > 0:
        if not _is_list_or_tuple_(max_sizes):
            max_sizes = [max_sizes]
        cur_max_sizes = max_sizes

    if in_dygraph_mode():
839 840 841 842 843 844
        attrs = ('min_sizes', min_sizes, 'aspect_ratios', aspect_ratios,
                 'variances', variance, 'flip', flip, 'clip', clip, 'step_w',
                 steps[0], 'step_h', steps[1], 'offset', offset,
                 'min_max_aspect_ratios_order', min_max_aspect_ratios_order)
        if cur_max_sizes is not None:
            attrs += ('max_sizes', cur_max_sizes)
Q
qingqing01 已提交
845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885
        box, var = core.ops.prior_box(input, image, *attrs)
        return box, var
    else:
        attrs = {
            'min_sizes': min_sizes,
            'aspect_ratios': aspect_ratios,
            'variances': variance,
            'flip': flip,
            'clip': clip,
            'step_w': steps[0],
            'step_h': steps[1],
            'offset': offset,
            'min_max_aspect_ratios_order': min_max_aspect_ratios_order
        }

        if cur_max_sizes is not None:
            attrs['max_sizes'] = cur_max_sizes

        box = helper.create_variable_for_type_inference(dtype)
        var = helper.create_variable_for_type_inference(dtype)
        helper.append_op(
            type="prior_box",
            inputs={"Input": input,
                    "Image": image},
            outputs={"Boxes": box,
                     "Variances": var},
            attrs=attrs, )
        box.stop_gradient = True
        var.stop_gradient = True
        return box, var


@paddle.jit.not_to_static
def multiclass_nms(bboxes,
                   scores,
                   score_threshold,
                   nms_top_k,
                   keep_top_k,
                   nms_threshold=0.3,
                   normalized=True,
                   nms_eta=1.,
886
                   background_label=-1,
Q
qingqing01 已提交
887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989
                   return_index=False,
                   return_rois_num=True,
                   rois_num=None,
                   name=None):
    """
    This operator is to do multi-class non maximum suppression (NMS) on
    boxes and scores.
    In the NMS step, this operator greedily selects a subset of detection bounding
    boxes that have high scores larger than score_threshold, if providing this
    threshold, then selects the largest nms_top_k confidences scores if nms_top_k
    is larger than -1. Then this operator pruns away boxes that have high IOU
    (intersection over union) overlap with already selected boxes by adaptive
    threshold NMS based on parameters of nms_threshold and nms_eta.
    Aftern NMS step, at most keep_top_k number of total bboxes are to be kept
    per image if keep_top_k is larger than -1.
    Args:
        bboxes (Tensor): Two types of bboxes are supported:
                           1. (Tensor) A 3-D Tensor with shape
                           [N, M, 4 or 8 16 24 32] represents the
                           predicted locations of M bounding bboxes,
                           N is the batch size. Each bounding box has four
                           coordinate values and the layout is
                           [xmin, ymin, xmax, ymax], when box size equals to 4.
                           2. (LoDTensor) A 3-D Tensor with shape [M, C, 4]
                           M is the number of bounding boxes, C is the
                           class number
        scores (Tensor): Two types of scores are supported:
                           1. (Tensor) A 3-D Tensor with shape [N, C, M]
                           represents the predicted confidence predictions.
                           N is the batch size, C is the class number, M is
                           number of bounding boxes. For each category there
                           are total M scores which corresponding M bounding
                           boxes. Please note, M is equal to the 2nd dimension
                           of BBoxes.
                           2. (LoDTensor) A 2-D LoDTensor with shape [M, C].
                           M is the number of bbox, C is the class number.
                           In this case, input BBoxes should be the second
                           case with shape [M, C, 4].
        background_label (int): The index of background label, the background
                                label will be ignored. If set to -1, then all
                                categories will be considered. Default: 0
        score_threshold (float): Threshold to filter out bounding boxes with
                                 low confidence score. If not provided,
                                 consider all boxes.
        nms_top_k (int): Maximum number of detections to be kept according to
                         the confidences after the filtering detections based
                         on score_threshold.
        nms_threshold (float): The threshold to be used in NMS. Default: 0.3
        nms_eta (float): The threshold to be used in NMS. Default: 1.0
        keep_top_k (int): Number of total bboxes to be kept per image after NMS
                          step. -1 means keeping all bboxes after NMS step.
        normalized (bool): Whether detections are normalized. Default: True
        return_index(bool): Whether return selected index. Default: False
        rois_num(Tensor): 1-D Tensor contains the number of RoIs in each image. 
            The shape is [B] and data type is int32. B is the number of images.
            If it is not None then return a list of 1-D Tensor. Each element 
            is the output RoIs' number of each image on the corresponding level
            and the shape is [B]. None by default.
        name(str): Name of the multiclass nms op. Default: None.
    Returns:
        A tuple with two Variables: (Out, Index) if return_index is True,
        otherwise, a tuple with one Variable(Out) is returned.
        Out: A 2-D LoDTensor with shape [No, 6] represents the detections.
        Each row has 6 values: [label, confidence, xmin, ymin, xmax, ymax]
        or A 2-D LoDTensor with shape [No, 10] represents the detections.
        Each row has 10 values: [label, confidence, x1, y1, x2, y2, x3, y3,
        x4, y4]. No is the total number of detections.
        If all images have not detected results, all elements in LoD will be
        0, and output tensor is empty (None).
        Index: Only return when return_index is True. A 2-D LoDTensor with
        shape [No, 1] represents the selected index which type is Integer.
        The index is the absolute value cross batches. No is the same number
        as Out. If the index is used to gather other attribute such as age,
        one needs to reshape the input(N, M, 1) to (N * M, 1) as first, where
        N is the batch size and M is the number of boxes.
    Examples:
        .. code-block:: python

            import paddle
            from ppdet.modeling import ops
            boxes = paddle.static.data(name='bboxes', shape=[81, 4],
                                      dtype='float32', lod_level=1)
            scores = paddle.static.data(name='scores', shape=[81],
                                      dtype='float32', lod_level=1)
            out, index = ops.multiclass_nms(bboxes=boxes,
                                            scores=scores,
                                            background_label=0,
                                            score_threshold=0.5,
                                            nms_top_k=400,
                                            nms_threshold=0.3,
                                            keep_top_k=200,
                                            normalized=False,
                                            return_index=True)
    """
    helper = LayerHelper('multiclass_nms3', **locals())

    if in_dygraph_mode():
        attrs = ('background_label', background_label, 'score_threshold',
                 score_threshold, 'nms_top_k', nms_top_k, 'nms_threshold',
                 nms_threshold, 'keep_top_k', keep_top_k, 'nms_eta', nms_eta,
                 'normalized', normalized)
        output, index, nms_rois_num = core.ops.multiclass_nms3(bboxes, scores,
                                                               rois_num, *attrs)
W
wangguanzhong 已提交
990
        if not return_index:
Q
qingqing01 已提交
991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130
            index = None
        return output, nms_rois_num, index

    else:
        output = helper.create_variable_for_type_inference(dtype=bboxes.dtype)
        index = helper.create_variable_for_type_inference(dtype='int')

        inputs = {'BBoxes': bboxes, 'Scores': scores}
        outputs = {'Out': output, 'Index': index}

        if rois_num is not None:
            inputs['RoisNum'] = rois_num

        if return_rois_num:
            nms_rois_num = helper.create_variable_for_type_inference(
                dtype='int32')
            outputs['NmsRoisNum'] = nms_rois_num

        helper.append_op(
            type="multiclass_nms3",
            inputs=inputs,
            attrs={
                'background_label': background_label,
                'score_threshold': score_threshold,
                'nms_top_k': nms_top_k,
                'nms_threshold': nms_threshold,
                'keep_top_k': keep_top_k,
                'nms_eta': nms_eta,
                'normalized': normalized
            },
            outputs=outputs)
        output.stop_gradient = True
        index.stop_gradient = True
        if not return_index:
            index = None
        if not return_rois_num:
            nms_rois_num = None

        return output, nms_rois_num, index


@paddle.jit.not_to_static
def matrix_nms(bboxes,
               scores,
               score_threshold,
               post_threshold,
               nms_top_k,
               keep_top_k,
               use_gaussian=False,
               gaussian_sigma=2.,
               background_label=0,
               normalized=True,
               return_index=False,
               return_rois_num=True,
               name=None):
    """
    **Matrix NMS**
    This operator does matrix non maximum suppression (NMS).
    First selects a subset of candidate bounding boxes that have higher scores
    than score_threshold (if provided), then the top k candidate is selected if
    nms_top_k is larger than -1. Score of the remaining candidate are then
    decayed according to the Matrix NMS scheme.
    Aftern NMS step, at most keep_top_k number of total bboxes are to be kept
    per image if keep_top_k is larger than -1.
    Args:
        bboxes (Tensor): A 3-D Tensor with shape [N, M, 4] represents the
                           predicted locations of M bounding bboxes,
                           N is the batch size. Each bounding box has four
                           coordinate values and the layout is
                           [xmin, ymin, xmax, ymax], when box size equals to 4.
                           The data type is float32 or float64.
        scores (Tensor): A 3-D Tensor with shape [N, C, M]
                           represents the predicted confidence predictions.
                           N is the batch size, C is the class number, M is
                           number of bounding boxes. For each category there
                           are total M scores which corresponding M bounding
                           boxes. Please note, M is equal to the 2nd dimension
                           of BBoxes. The data type is float32 or float64.
        score_threshold (float): Threshold to filter out bounding boxes with
                                 low confidence score.
        post_threshold (float): Threshold to filter out bounding boxes with
                                low confidence score AFTER decaying.
        nms_top_k (int): Maximum number of detections to be kept according to
                         the confidences after the filtering detections based
                         on score_threshold.
        keep_top_k (int): Number of total bboxes to be kept per image after NMS
                          step. -1 means keeping all bboxes after NMS step.
        use_gaussian (bool): Use Gaussian as the decay function. Default: False
        gaussian_sigma (float): Sigma for Gaussian decay function. Default: 2.0
        background_label (int): The index of background label, the background
                                label will be ignored. If set to -1, then all
                                categories will be considered. Default: 0
        normalized (bool): Whether detections are normalized. Default: True
        return_index(bool): Whether return selected index. Default: False
        return_rois_num(bool): whether return rois_num. Default: True
        name(str): Name of the matrix nms op. Default: None.
    Returns:
        A tuple with three Tensor: (Out, Index, RoisNum) if return_index is True,
        otherwise, a tuple with two Tensor (Out, RoisNum) is returned.
        Out (Tensor): A 2-D Tensor with shape [No, 6] containing the
             detection results.
             Each row has 6 values: [label, confidence, xmin, ymin, xmax, ymax]
             (After version 1.3, when no boxes detected, the lod is changed
             from {0} to {1})
        Index (Tensor): A 2-D Tensor with shape [No, 1] containing the
            selected indices, which are absolute values cross batches.
        rois_num (Tensor): A 1-D Tensor with shape [N] containing 
            the number of detected boxes in each image.
    Examples:
        .. code-block:: python
            import paddle
            from ppdet.modeling import ops
            boxes = paddle.static.data(name='bboxes', shape=[None,81, 4],
                                      dtype='float32', lod_level=1)
            scores = paddle.static.data(name='scores', shape=[None,81],
                                      dtype='float32', lod_level=1)
            out = ops.matrix_nms(bboxes=boxes, scores=scores, background_label=0,
                                 score_threshold=0.5, post_threshold=0.1,
                                 nms_top_k=400, keep_top_k=200, normalized=False)
    """
    check_variable_and_dtype(bboxes, 'BBoxes', ['float32', 'float64'],
                             'matrix_nms')
    check_variable_and_dtype(scores, 'Scores', ['float32', 'float64'],
                             'matrix_nms')
    check_type(score_threshold, 'score_threshold', float, 'matrix_nms')
    check_type(post_threshold, 'post_threshold', float, 'matrix_nms')
    check_type(nms_top_k, 'nums_top_k', int, 'matrix_nms')
    check_type(keep_top_k, 'keep_top_k', int, 'matrix_nms')
    check_type(normalized, 'normalized', bool, 'matrix_nms')
    check_type(use_gaussian, 'use_gaussian', bool, 'matrix_nms')
    check_type(gaussian_sigma, 'gaussian_sigma', float, 'matrix_nms')
    check_type(background_label, 'background_label', int, 'matrix_nms')

    if in_dygraph_mode():
        attrs = ('background_label', background_label, 'score_threshold',
                 score_threshold, 'post_threshold', post_threshold, 'nms_top_k',
                 nms_top_k, 'gaussian_sigma', gaussian_sigma, 'use_gaussian',
                 use_gaussian, 'keep_top_k', keep_top_k, 'normalized',
                 normalized)
        out, index, rois_num = core.ops.matrix_nms(bboxes, scores, *attrs)
W
wangxinxin08 已提交
1131 1132 1133 1134 1135
        if not return_index:
            index = None
        if not return_rois_num:
            rois_num = None
        return out, rois_num, index
Q
qingqing01 已提交
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161
    else:
        helper = LayerHelper('matrix_nms', **locals())
        output = helper.create_variable_for_type_inference(dtype=bboxes.dtype)
        index = helper.create_variable_for_type_inference(dtype='int')
        outputs = {'Out': output, 'Index': index}
        if return_rois_num:
            rois_num = helper.create_variable_for_type_inference(dtype='int')
            outputs['RoisNum'] = rois_num

        helper.append_op(
            type="matrix_nms",
            inputs={'BBoxes': bboxes,
                    'Scores': scores},
            attrs={
                'background_label': background_label,
                'score_threshold': score_threshold,
                'post_threshold': post_threshold,
                'nms_top_k': nms_top_k,
                'gaussian_sigma': gaussian_sigma,
                'use_gaussian': use_gaussian,
                'keep_top_k': keep_top_k,
                'normalized': normalized
            },
            outputs=outputs)
        output.stop_gradient = True

W
wangxinxin08 已提交
1162 1163 1164 1165 1166
        if not return_index:
            index = None
        if not return_rois_num:
            rois_num = None
        return output, rois_num, index
Q
qingqing01 已提交
1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438


def bipartite_match(dist_matrix,
                    match_type=None,
                    dist_threshold=None,
                    name=None):
    """

    This operator implements a greedy bipartite matching algorithm, which is
    used to obtain the matching with the maximum distance based on the input
    distance matrix. For input 2D matrix, the bipartite matching algorithm can
    find the matched column for each row (matched means the largest distance),
    also can find the matched row for each column. And this operator only
    calculate matched indices from column to row. For each instance,
    the number of matched indices is the column number of the input distance
    matrix. **The OP only supports CPU**.

    There are two outputs, matched indices and distance.
    A simple description, this algorithm matched the best (maximum distance)
    row entity to the column entity and the matched indices are not duplicated
    in each row of ColToRowMatchIndices. If the column entity is not matched
    any row entity, set -1 in ColToRowMatchIndices.

    NOTE: the input DistMat can be LoDTensor (with LoD) or Tensor.
    If LoDTensor with LoD, the height of ColToRowMatchIndices is batch size.
    If Tensor, the height of ColToRowMatchIndices is 1.

    NOTE: This API is a very low level API. It is used by :code:`ssd_loss`
    layer. Please consider to use :code:`ssd_loss` instead.

    Args:
        dist_matrix(Tensor): This input is a 2-D LoDTensor with shape
            [K, M]. The data type is float32 or float64. It is pair-wise 
            distance matrix between the entities represented by each row and 
            each column. For example, assumed one entity is A with shape [K], 
            another entity is B with shape [M]. The dist_matrix[i][j] is the 
            distance between A[i] and B[j]. The bigger the distance is, the 
            better matching the pairs are. NOTE: This tensor can contain LoD 
            information to represent a batch of inputs. One instance of this 
            batch can contain different numbers of entities.
        match_type(str, optional): The type of matching method, should be
           'bipartite' or 'per_prediction'. None ('bipartite') by default.
        dist_threshold(float32, optional): If `match_type` is 'per_prediction',
            this threshold is to determine the extra matching bboxes based
            on the maximum distance, 0.5 by default.
        name(str, optional): For detailed information, please refer 
            to :ref:`api_guide_Name`. Usually name is no need to set and 
            None by default.
 
    Returns:
        Tuple:

        matched_indices(Tensor): A 2-D Tensor with shape [N, M]. The data
        type is int32. N is the batch size. If match_indices[i][j] is -1, it
        means B[j] does not match any entity in i-th instance.
        Otherwise, it means B[j] is matched to row
        match_indices[i][j] in i-th instance. The row number of
        i-th instance is saved in match_indices[i][j].

        matched_distance(Tensor): A 2-D Tensor with shape [N, M]. The data
        type is float32. N is batch size. If match_indices[i][j] is -1,
        match_distance[i][j] is also -1.0. Otherwise, assumed
        match_distance[i][j] = d, and the row offsets of each instance
        are called LoD. Then match_distance[i][j] =
        dist_matrix[d+LoD[i]][j].

    Examples:

        .. code-block:: python
            import paddle
            from ppdet.modeling import ops
            from ppdet.modeling.utils import iou_similarity

            paddle.enable_static()

            x = paddle.static.data(name='x', shape=[None, 4], dtype='float32')
            y = paddle.static.data(name='y', shape=[None, 4], dtype='float32')
            iou = iou_similarity(x=x, y=y)
            matched_indices, matched_dist = ops.bipartite_match(iou)
    """
    check_variable_and_dtype(dist_matrix, 'dist_matrix',
                             ['float32', 'float64'], 'bipartite_match')

    if in_dygraph_mode():
        match_indices, match_distance = core.ops.bipartite_match(
            dist_matrix, "match_type", match_type, "dist_threshold",
            dist_threshold)
        return match_indices, match_distance

    helper = LayerHelper('bipartite_match', **locals())
    match_indices = helper.create_variable_for_type_inference(dtype='int32')
    match_distance = helper.create_variable_for_type_inference(
        dtype=dist_matrix.dtype)
    helper.append_op(
        type='bipartite_match',
        inputs={'DistMat': dist_matrix},
        attrs={
            'match_type': match_type,
            'dist_threshold': dist_threshold,
        },
        outputs={
            'ColToRowMatchIndices': match_indices,
            'ColToRowMatchDist': match_distance
        })
    return match_indices, match_distance


@paddle.jit.not_to_static
def box_coder(prior_box,
              prior_box_var,
              target_box,
              code_type="encode_center_size",
              box_normalized=True,
              axis=0,
              name=None):
    """
    **Box Coder Layer**
    Encode/Decode the target bounding box with the priorbox information.
    
    The Encoding schema described below:
    .. math::
        ox = (tx - px) / pw / pxv
        oy = (ty - py) / ph / pyv
        ow = \log(\abs(tw / pw)) / pwv 
        oh = \log(\abs(th / ph)) / phv 
    The Decoding schema described below:
    
    .. math::
  
        ox = (pw * pxv * tx * + px) - tw / 2
        oy = (ph * pyv * ty * + py) - th / 2
        ow = \exp(pwv * tw) * pw + tw / 2
        oh = \exp(phv * th) * ph + th / 2   
    where `tx`, `ty`, `tw`, `th` denote the target box's center coordinates, 
    width and height respectively. Similarly, `px`, `py`, `pw`, `ph` denote 
    the priorbox's (anchor) center coordinates, width and height. `pxv`, 
    `pyv`, `pwv`, `phv` denote the variance of the priorbox and `ox`, `oy`, 
    `ow`, `oh` denote the encoded/decoded coordinates, width and height. 
    During Box Decoding, two modes for broadcast are supported. Say target 
    box has shape [N, M, 4], and the shape of prior box can be [N, 4] or 
    [M, 4]. Then prior box will broadcast to target box along the 
    assigned axis. 

    Args:
        prior_box(Tensor): Box list prior_box is a 2-D Tensor with shape 
            [M, 4] holds M boxes and data type is float32 or float64. Each box
            is represented as [xmin, ymin, xmax, ymax], [xmin, ymin] is the 
            left top coordinate of the anchor box, if the input is image feature
            map, they are close to the origin of the coordinate system. 
            [xmax, ymax] is the right bottom coordinate of the anchor box.       
        prior_box_var(List|Tensor|None): prior_box_var supports three types 
            of input. One is Tensor with shape [M, 4] which holds M group and 
            data type is float32 or float64. The second is list consist of 
            4 elements shared by all boxes and data type is float32 or float64. 
            Other is None and not involved in calculation. 
        target_box(Tensor): This input can be a 2-D LoDTensor with shape 
            [N, 4] when code_type is 'encode_center_size'. This input also can 
            be a 3-D Tensor with shape [N, M, 4] when code_type is 
            'decode_center_size'. Each box is represented as 
            [xmin, ymin, xmax, ymax]. The data type is float32 or float64. 
        code_type(str): The code type used with the target box. It can be
            `encode_center_size` or `decode_center_size`. `encode_center_size` 
            by default.
        box_normalized(bool): Whether treat the priorbox as a normalized box.
            Set true by default.
        axis(int): Which axis in PriorBox to broadcast for box decode, 
            for example, if axis is 0 and TargetBox has shape [N, M, 4] and 
            PriorBox has shape [M, 4], then PriorBox will broadcast to [N, M, 4]
            for decoding. It is only valid when code type is 
            `decode_center_size`. Set 0 by default. 
        name(str, optional): For detailed information, please refer 
            to :ref:`api_guide_Name`. Usually name is no need to set and 
            None by default. 

    Returns:
        Tensor:
        output_box(Tensor): When code_type is 'encode_center_size', the 
        output tensor of box_coder_op with shape [N, M, 4] representing the 
        result of N target boxes encoded with M Prior boxes and variances. 
        When code_type is 'decode_center_size', N represents the batch size 
        and M represents the number of decoded boxes.

    Examples:
 
        .. code-block:: python
 
            import paddle
            from ppdet.modeling import ops
            paddle.enable_static()
            # For encode
            prior_box_encode = paddle.static.data(name='prior_box_encode',
                                  shape=[512, 4],
                                  dtype='float32')
            target_box_encode = paddle.static.data(name='target_box_encode',
                                   shape=[81, 4],
                                   dtype='float32')
            output_encode = ops.box_coder(prior_box=prior_box_encode,
                                    prior_box_var=[0.1,0.1,0.2,0.2],
                                    target_box=target_box_encode,
                                    code_type="encode_center_size")
            # For decode
            prior_box_decode = paddle.static.data(name='prior_box_decode',
                                  shape=[512, 4],
                                  dtype='float32')
            target_box_decode = paddle.static.data(name='target_box_decode',
                                   shape=[512, 81, 4],
                                   dtype='float32')
            output_decode = ops.box_coder(prior_box=prior_box_decode,
                                    prior_box_var=[0.1,0.1,0.2,0.2],
                                    target_box=target_box_decode,
                                    code_type="decode_center_size",
                                    box_normalized=False,
                                    axis=1)
    """
    check_variable_and_dtype(prior_box, 'prior_box', ['float32', 'float64'],
                             'box_coder')
    check_variable_and_dtype(target_box, 'target_box', ['float32', 'float64'],
                             'box_coder')

    if in_dygraph_mode():
        if isinstance(prior_box_var, Variable):
            output_box = core.ops.box_coder(
                prior_box, prior_box_var, target_box, "code_type", code_type,
                "box_normalized", box_normalized, "axis", axis)

        elif isinstance(prior_box_var, list):
            output_box = core.ops.box_coder(
                prior_box, None, target_box, "code_type", code_type,
                "box_normalized", box_normalized, "axis", axis, "variance",
                prior_box_var)
        else:
            raise TypeError(
                "Input variance of box_coder must be Variable or list")
        return output_box
    else:
        helper = LayerHelper("box_coder", **locals())

        output_box = helper.create_variable_for_type_inference(
            dtype=prior_box.dtype)

        inputs = {"PriorBox": prior_box, "TargetBox": target_box}
        attrs = {
            "code_type": code_type,
            "box_normalized": box_normalized,
            "axis": axis
        }
        if isinstance(prior_box_var, Variable):
            inputs['PriorBoxVar'] = prior_box_var
        elif isinstance(prior_box_var, list):
            attrs['variance'] = prior_box_var
        else:
            raise TypeError(
                "Input variance of box_coder must be Variable or list")
        helper.append_op(
            type="box_coder",
            inputs=inputs,
            attrs=attrs,
            outputs={"OutputBox": output_box})
        return output_box


@paddle.jit.not_to_static
def generate_proposals(scores,
                       bbox_deltas,
                       im_shape,
                       anchors,
                       variances,
                       pre_nms_top_n=6000,
                       post_nms_top_n=1000,
                       nms_thresh=0.5,
                       min_size=0.1,
                       eta=1.0,
1439
                       pixel_offset=False,
Q
qingqing01 已提交
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513
                       return_rois_num=False,
                       name=None):
    """
    **Generate proposal Faster-RCNN**
    This operation proposes RoIs according to each box with their
    probability to be a foreground object and 
    the box can be calculated by anchors. Bbox_deltais and scores
    to be an object are the output of RPN. Final proposals
    could be used to train detection net.
    For generating proposals, this operation performs following steps:
    1. Transposes and resizes scores and bbox_deltas in size of
       (H*W*A, 1) and (H*W*A, 4)
    2. Calculate box locations as proposals candidates. 
    3. Clip boxes to image
    4. Remove predicted boxes with small area. 
    5. Apply NMS to get final proposals as output.
    Args:
        scores(Tensor): A 4-D Tensor with shape [N, A, H, W] represents
            the probability for each box to be an object.
            N is batch size, A is number of anchors, H and W are height and
            width of the feature map. The data type must be float32.
        bbox_deltas(Tensor): A 4-D Tensor with shape [N, 4*A, H, W]
            represents the difference between predicted box location and
            anchor location. The data type must be float32.
        im_shape(Tensor): A 2-D Tensor with shape [N, 2] represents H, W, the
            origin image size or input size. The data type can be float32 or 
            float64.
        anchors(Tensor):   A 4-D Tensor represents the anchors with a layout
            of [H, W, A, 4]. H and W are height and width of the feature map,
            num_anchors is the box count of each position. Each anchor is
            in (xmin, ymin, xmax, ymax) format an unnormalized. The data type must be float32.
        variances(Tensor): A 4-D Tensor. The expanded variances of anchors with a layout of
            [H, W, num_priors, 4]. Each variance is in
            (xcenter, ycenter, w, h) format. The data type must be float32.
        pre_nms_top_n(float): Number of total bboxes to be kept per
            image before NMS. The data type must be float32. `6000` by default.
        post_nms_top_n(float): Number of total bboxes to be kept per
            image after NMS. The data type must be float32. `1000` by default.
        nms_thresh(float): Threshold in NMS. The data type must be float32. `0.5` by default.
        min_size(float): Remove predicted boxes with either height or
            width < min_size. The data type must be float32. `0.1` by default.
        eta(float): Apply in adaptive NMS, if adaptive `threshold > 0.5`,
            `adaptive_threshold = adaptive_threshold * eta` in each iteration.
        return_rois_num(bool): When setting True, it will return a 1D Tensor with shape [N, ] that includes Rois's 
            num of each image in one batch. The N is the image's num. For example, the tensor has values [4,5] that represents
            the first image has 4 Rois, the second image has 5 Rois. It only used in rcnn model. 
            'False' by default. 
        name(str, optional): For detailed information, please refer 
            to :ref:`api_guide_Name`. Usually name is no need to set and 
            None by default. 

    Returns:
        tuple:
        A tuple with format ``(rpn_rois, rpn_roi_probs)``.
        - **rpn_rois**: The generated RoIs. 2-D Tensor with shape ``[N, 4]`` while ``N`` is the number of RoIs. The data type is the same as ``scores``.
        - **rpn_roi_probs**: The scores of generated RoIs. 2-D Tensor with shape ``[N, 1]`` while ``N`` is the number of RoIs. The data type is the same as ``scores``.

    Examples:
        .. code-block:: python
        
            import paddle
            from ppdet.modeling import ops
            paddle.enable_static()
            scores = paddle.static.data(name='scores', shape=[None, 4, 5, 5], dtype='float32')
            bbox_deltas = paddle.static.data(name='bbox_deltas', shape=[None, 16, 5, 5], dtype='float32')
            im_shape = paddle.static.data(name='im_shape', shape=[None, 2], dtype='float32')
            anchors = paddle.static.data(name='anchors', shape=[None, 5, 4, 4], dtype='float32')
            variances = paddle.static.data(name='variances', shape=[None, 5, 10, 4], dtype='float32')
            rois, roi_probs = ops.generate_proposals(scores, bbox_deltas,
                         im_shape, anchors, variances)
    """
    if in_dygraph_mode():
        assert return_rois_num, "return_rois_num should be True in dygraph mode."
        attrs = ('pre_nms_topN', pre_nms_top_n, 'post_nms_topN', post_nms_top_n,
1514 1515
                 'nms_thresh', nms_thresh, 'min_size', min_size, 'eta', eta,
                 'pixel_offset', pixel_offset)
Q
qingqing01 已提交
1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
        rpn_rois, rpn_roi_probs, rpn_rois_num = core.ops.generate_proposals_v2(
            scores, bbox_deltas, im_shape, anchors, variances, *attrs)
        return rpn_rois, rpn_roi_probs, rpn_rois_num

    else:
        helper = LayerHelper('generate_proposals_v2', **locals())

        check_variable_and_dtype(scores, 'scores', ['float32'],
                                 'generate_proposals_v2')
        check_variable_and_dtype(bbox_deltas, 'bbox_deltas', ['float32'],
                                 'generate_proposals_v2')
        check_variable_and_dtype(im_shape, 'im_shape', ['float32', 'float64'],
                                 'generate_proposals_v2')
        check_variable_and_dtype(anchors, 'anchors', ['float32'],
                                 'generate_proposals_v2')
        check_variable_and_dtype(variances, 'variances', ['float32'],
                                 'generate_proposals_v2')

        rpn_rois = helper.create_variable_for_type_inference(
            dtype=bbox_deltas.dtype)
        rpn_roi_probs = helper.create_variable_for_type_inference(
            dtype=scores.dtype)
        outputs = {
            'RpnRois': rpn_rois,
            'RpnRoiProbs': rpn_roi_probs,
        }
        if return_rois_num:
            rpn_rois_num = helper.create_variable_for_type_inference(
                dtype='int32')
            rpn_rois_num.stop_gradient = True
            outputs['RpnRoisNum'] = rpn_rois_num

        helper.append_op(
            type="generate_proposals_v2",
            inputs={
                'Scores': scores,
                'BboxDeltas': bbox_deltas,
                'ImShape': im_shape,
                'Anchors': anchors,
                'Variances': variances
            },
            attrs={
                'pre_nms_topN': pre_nms_top_n,
                'post_nms_topN': post_nms_top_n,
                'nms_thresh': nms_thresh,
                'min_size': min_size,
1562 1563
                'eta': eta,
                'pixel_offset': pixel_offset
Q
qingqing01 已提交
1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595
            },
            outputs=outputs)
        rpn_rois.stop_gradient = True
        rpn_roi_probs.stop_gradient = True

        return rpn_rois, rpn_roi_probs, rpn_rois_num


def sigmoid_cross_entropy_with_logits(input,
                                      label,
                                      ignore_index=-100,
                                      normalize=False):
    output = F.binary_cross_entropy_with_logits(input, label, reduction='none')
    mask_tensor = paddle.cast(label != ignore_index, 'float32')
    output = paddle.multiply(output, mask_tensor)
    if normalize:
        sum_valid_mask = paddle.sum(mask_tensor)
        output = output / sum_valid_mask
    return output


def smooth_l1(input, label, inside_weight=None, outside_weight=None,
              sigma=None):
    input_new = paddle.multiply(input, inside_weight)
    label_new = paddle.multiply(label, inside_weight)
    delta = 1 / (sigma * sigma)
    out = F.smooth_l1_loss(input_new, label_new, reduction='none', delta=delta)
    out = paddle.multiply(out, outside_weight)
    out = out / delta
    out = paddle.reshape(out, shape=[out.shape[0], -1])
    out = paddle.sum(out, axis=1)
    return out