detr_loss.py 13.1 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
# Copyright (c) 2021 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.

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import paddle
import paddle.nn as nn
import paddle.nn.functional as F
from ppdet.core.workspace import register
from .iou_loss import GIoULoss
S
shangliang Xu 已提交
24
from ..transformers import bbox_cxcywh_to_xyxy, sigmoid_focal_loss
25

S
shangliang Xu 已提交
26
__all__ = ['DETRLoss', 'DINOLoss']
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69


@register
class DETRLoss(nn.Layer):
    __shared__ = ['num_classes', 'use_focal_loss']
    __inject__ = ['matcher']

    def __init__(self,
                 num_classes=80,
                 matcher='HungarianMatcher',
                 loss_coeff={
                     'class': 1,
                     'bbox': 5,
                     'giou': 2,
                     'no_object': 0.1,
                     'mask': 1,
                     'dice': 1
                 },
                 aux_loss=True,
                 use_focal_loss=False):
        r"""
        Args:
            num_classes (int): The number of classes.
            matcher (HungarianMatcher): It computes an assignment between the targets
                and the predictions of the network.
            loss_coeff (dict): The coefficient of loss.
            aux_loss (bool): If 'aux_loss = True', loss at each decoder layer are to be used.
            use_focal_loss (bool): Use focal loss or not.
        """
        super(DETRLoss, self).__init__()
        self.num_classes = num_classes

        self.matcher = matcher
        self.loss_coeff = loss_coeff
        self.aux_loss = aux_loss
        self.use_focal_loss = use_focal_loss

        if not self.use_focal_loss:
            self.loss_coeff['class'] = paddle.full([num_classes + 1],
                                                   loss_coeff['class'])
            self.loss_coeff['class'][-1] = loss_coeff['no_object']
        self.giou_loss = GIoULoss()

S
shangliang Xu 已提交
70 71 72 73 74 75 76
    def _get_loss_class(self,
                        logits,
                        gt_class,
                        match_indices,
                        bg_index,
                        num_gts,
                        postfix=""):
77
        # logits: [b, query, num_classes], gt_class: list[[n, 1]]
S
shangliang Xu 已提交
78 79 80
        name_class = "loss_class" + postfix
        if logits is None:
            return {name_class: paddle.zeros([1])}
81 82 83 84 85 86 87 88 89 90
        target_label = paddle.full(logits.shape[:2], bg_index, dtype='int64')
        bs, num_query_objects = target_label.shape
        if sum(len(a) for a in gt_class) > 0:
            index, updates = self._get_index_updates(num_query_objects,
                                                     gt_class, match_indices)
            target_label = paddle.scatter(
                target_label.reshape([-1, 1]), index, updates.astype('int64'))
            target_label = target_label.reshape([bs, num_query_objects])
        if self.use_focal_loss:
            target_label = F.one_hot(target_label,
S
shangliang Xu 已提交
91
                                     self.num_classes + 1)[..., :-1]
92
        return {
S
shangliang Xu 已提交
93
            name_class: self.loss_coeff['class'] * sigmoid_focal_loss(
94 95 96 97 98
                logits, target_label, num_gts / num_query_objects)
            if self.use_focal_loss else F.cross_entropy(
                logits, target_label, weight=self.loss_coeff['class'])
        }

S
shangliang Xu 已提交
99 100
    def _get_loss_bbox(self, boxes, gt_bbox, match_indices, num_gts,
                       postfix=""):
101
        # boxes: [b, query, 4], gt_bbox: list[[n, 4]]
S
shangliang Xu 已提交
102 103 104 105
        name_bbox = "loss_bbox" + postfix
        name_giou = "loss_giou" + postfix
        if boxes is None:
            return {name_bbox: paddle.zeros([1]), name_giou: paddle.zeros([1])}
106 107
        loss = dict()
        if sum(len(a) for a in gt_bbox) == 0:
S
shangliang Xu 已提交
108 109
            loss[name_bbox] = paddle.to_tensor([0.])
            loss[name_giou] = paddle.to_tensor([0.])
110 111 112 113
            return loss

        src_bbox, target_bbox = self._get_src_target_assign(boxes, gt_bbox,
                                                            match_indices)
S
shangliang Xu 已提交
114
        loss[name_bbox] = self.loss_coeff['bbox'] * F.l1_loss(
115
            src_bbox, target_bbox, reduction='sum') / num_gts
S
shangliang Xu 已提交
116
        loss[name_giou] = self.giou_loss(
117
            bbox_cxcywh_to_xyxy(src_bbox), bbox_cxcywh_to_xyxy(target_bbox))
S
shangliang Xu 已提交
118 119
        loss[name_giou] = loss[name_giou].sum() / num_gts
        loss[name_giou] = self.loss_coeff['giou'] * loss[name_giou]
120 121
        return loss

S
shangliang Xu 已提交
122 123
    def _get_loss_mask(self, masks, gt_mask, match_indices, num_gts,
                       postfix=""):
124
        # masks: [b, query, h, w], gt_mask: list[[n, H, W]]
S
shangliang Xu 已提交
125 126 127 128
        name_mask = "loss_mask" + postfix
        name_dice = "loss_dice" + postfix
        if masks is None:
            return {name_mask: paddle.zeros([1]), name_dice: paddle.zeros([1])}
129 130
        loss = dict()
        if sum(len(a) for a in gt_mask) == 0:
S
shangliang Xu 已提交
131 132
            loss[name_mask] = paddle.to_tensor([0.])
            loss[name_dice] = paddle.to_tensor([0.])
133 134 135 136 137 138 139 140
            return loss

        src_masks, target_masks = self._get_src_target_assign(masks, gt_mask,
                                                              match_indices)
        src_masks = F.interpolate(
            src_masks.unsqueeze(0),
            size=target_masks.shape[-2:],
            mode="bilinear")[0]
S
shangliang Xu 已提交
141
        loss[name_mask] = self.loss_coeff['mask'] * F.sigmoid_focal_loss(
142 143 144 145
            src_masks,
            target_masks,
            paddle.to_tensor(
                [num_gts], dtype='float32'))
S
shangliang Xu 已提交
146
        loss[name_dice] = self.loss_coeff['dice'] * self._dice_loss(
147 148 149 150 151 152 153 154 155 156 157 158
            src_masks, target_masks, num_gts)
        return loss

    def _dice_loss(self, inputs, targets, num_gts):
        inputs = F.sigmoid(inputs)
        inputs = inputs.flatten(1)
        targets = targets.flatten(1)
        numerator = 2 * (inputs * targets).sum(1)
        denominator = inputs.sum(-1) + targets.sum(-1)
        loss = 1 - (numerator + 1) / (denominator + 1)
        return loss.sum() / num_gts

S
shangliang Xu 已提交
159 160 161 162 163 164 165
    def _get_loss_aux(self,
                      boxes,
                      logits,
                      gt_bbox,
                      gt_class,
                      bg_index,
                      num_gts,
166
                      dn_match_indices=None,
S
shangliang Xu 已提交
167
                      postfix=""):
168
        if boxes is None or logits is None:
S
shangliang Xu 已提交
169 170 171 172 173
            return {
                "loss_class_aux" + postfix: paddle.paddle.zeros([1]),
                "loss_bbox_aux" + postfix: paddle.paddle.zeros([1]),
                "loss_giou_aux" + postfix: paddle.paddle.zeros([1])
            }
174 175 176 177
        loss_class = []
        loss_bbox = []
        loss_giou = []
        for aux_boxes, aux_logits in zip(boxes, logits):
178
            if dn_match_indices is None:
S
shangliang Xu 已提交
179 180
                match_indices = self.matcher(aux_boxes, aux_logits, gt_bbox,
                                             gt_class)
181 182
            else:
                match_indices = dn_match_indices
183 184
            loss_class.append(
                self._get_loss_class(aux_logits, gt_class, match_indices,
S
shangliang Xu 已提交
185 186
                                     bg_index, num_gts, postfix)['loss_class' +
                                                                 postfix])
187
            loss_ = self._get_loss_bbox(aux_boxes, gt_bbox, match_indices,
S
shangliang Xu 已提交
188 189 190
                                        num_gts, postfix)
            loss_bbox.append(loss_['loss_bbox' + postfix])
            loss_giou.append(loss_['loss_giou' + postfix])
191
        loss = {
S
shangliang Xu 已提交
192 193 194
            "loss_class_aux" + postfix: paddle.add_n(loss_class),
            "loss_bbox_aux" + postfix: paddle.add_n(loss_bbox),
            "loss_giou_aux" + postfix: paddle.add_n(loss_giou)
195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
        }
        return loss

    def _get_index_updates(self, num_query_objects, target, match_indices):
        batch_idx = paddle.concat([
            paddle.full_like(src, i) for i, (src, _) in enumerate(match_indices)
        ])
        src_idx = paddle.concat([src for (src, _) in match_indices])
        src_idx += (batch_idx * num_query_objects)
        target_assign = paddle.concat([
            paddle.gather(
                t, dst, axis=0) for t, (_, dst) in zip(target, match_indices)
        ])
        return src_idx, target_assign

    def _get_src_target_assign(self, src, target, match_indices):
        src_assign = paddle.concat([
            paddle.gather(
                t, I, axis=0) if len(I) > 0 else paddle.zeros([0, t.shape[-1]])
            for t, (I, _) in zip(src, match_indices)
        ])
        target_assign = paddle.concat([
            paddle.gather(
                t, J, axis=0) if len(J) > 0 else paddle.zeros([0, t.shape[-1]])
            for t, (_, J) in zip(target, match_indices)
        ])
        return src_assign, target_assign

    def forward(self,
                boxes,
                logits,
                gt_bbox,
                gt_class,
                masks=None,
S
shangliang Xu 已提交
229 230 231
                gt_mask=None,
                postfix="",
                **kwargs):
232 233
        r"""
        Args:
S
shangliang Xu 已提交
234 235
            boxes (Tensor|None): [l, b, query, 4]
            logits (Tensor|None): [l, b, query, num_classes]
236 237 238 239
            gt_bbox (List(Tensor)): list[[n, 4]]
            gt_class (List(Tensor)): list[[n, 1]]
            masks (Tensor, optional): [b, query, h, w]
            gt_mask (List(Tensor), optional): list[[n, H, W]]
S
shangliang Xu 已提交
240
            postfix (str): postfix of loss name
241
        """
242 243 244
        dn_match_indices = kwargs.get("dn_match_indices", None)
        if dn_match_indices is None and (boxes is not None and
                                         logits is not None):
S
shangliang Xu 已提交
245 246
            match_indices = self.matcher(boxes[-1].detach(),
                                         logits[-1].detach(), gt_bbox, gt_class)
247 248
        else:
            match_indices = dn_match_indices
S
shangliang Xu 已提交
249

250
        num_gts = sum(len(a) for a in gt_bbox)
S
shangliang Xu 已提交
251 252
        num_gts = paddle.to_tensor([num_gts], dtype="float32")
        if paddle.distributed.get_world_size() > 1:
253
            paddle.distributed.all_reduce(num_gts)
S
shangliang Xu 已提交
254 255 256
            num_gts /= paddle.distributed.get_world_size()
        num_gts = paddle.clip(num_gts, min=1.) * kwargs.get("dn_num_group", 1.)

257 258
        total_loss = dict()
        total_loss.update(
S
shangliang Xu 已提交
259 260 261
            self._get_loss_class(logits[
                -1] if logits is not None else None, gt_class, match_indices,
                                 self.num_classes, num_gts, postfix))
262
        total_loss.update(
S
shangliang Xu 已提交
263 264
            self._get_loss_bbox(boxes[-1] if boxes is not None else None,
                                gt_bbox, match_indices, num_gts, postfix))
265 266
        if masks is not None and gt_mask is not None:
            total_loss.update(
S
shangliang Xu 已提交
267 268
                self._get_loss_mask(masks if masks is not None else None,
                                    gt_mask, match_indices, num_gts, postfix))
269 270 271

        if self.aux_loss:
            total_loss.update(
S
shangliang Xu 已提交
272 273 274
                self._get_loss_aux(
                    boxes[:-1] if boxes is not None else None, logits[:-1]
                    if logits is not None else None, gt_bbox, gt_class,
275
                    self.num_classes, num_gts, dn_match_indices, postfix))
S
shangliang Xu 已提交
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318

        return total_loss


@register
class DINOLoss(DETRLoss):
    def forward(self,
                boxes,
                logits,
                gt_bbox,
                gt_class,
                masks=None,
                gt_mask=None,
                postfix="",
                dn_out_bboxes=None,
                dn_out_logits=None,
                dn_meta=None,
                **kwargs):
        total_loss = super(DINOLoss, self).forward(boxes, logits, gt_bbox,
                                                   gt_class)

        if dn_meta is not None:
            dn_positive_idx, dn_num_group = \
                dn_meta["dn_positive_idx"], dn_meta["dn_num_group"]
            assert len(gt_class) == len(dn_positive_idx)

            # denoising match indices
            dn_match_indices = []
            for i in range(len(gt_class)):
                num_gt = len(gt_class[i])
                if num_gt > 0:
                    gt_idx = paddle.arange(end=num_gt, dtype="int64")
                    gt_idx = gt_idx.unsqueeze(0).tile(
                        [dn_num_group, 1]).flatten()
                    assert len(gt_idx) == len(dn_positive_idx[i])
                    dn_match_indices.append((dn_positive_idx[i], gt_idx))
                else:
                    dn_match_indices.append((paddle.zeros(
                        [0], dtype="int64"), paddle.zeros(
                            [0], dtype="int64")))
        else:
            dn_match_indices, dn_num_group = None, 1.

319
        # compute denoising training loss
S
shangliang Xu 已提交
320 321 322 323 324 325
        dn_loss = super(DINOLoss, self).forward(
            dn_out_bboxes,
            dn_out_logits,
            gt_bbox,
            gt_class,
            postfix="_dn",
326
            dn_match_indices=dn_match_indices,
S
shangliang Xu 已提交
327 328
            dn_num_group=dn_num_group)
        total_loss.update(dn_loss)
329 330

        return total_loss