Add ssd (#1782)

* add ssd_vgg16_120e_coco

configs/_base_/datasets/voc.yml

+metric: VOC
+num_classes: 20
+
+TrainDataset:
+  !VOCDataSet
+    dataset_dir: dataset/voc
+    anno_path: trainval.txt
+    label_list: label_list.txt
+
+EvalDataset:
+  !VOCDataSet
+    dataset_dir: dataset/voc
+    anno_path: test.txt
+    label_list: label_list.txt
+
+TestDataset:
+  !ImageFolder
+    anno_path: dataset/voc/label_list.txt

configs/_base_/models/ssd_vgg16_300.yml

+architecture: SSD
+pretrain_weights: https://paddlemodels.bj.bcebos.com/object_detection/dygraph/VGG16_caffe_pretrained.pdparams
+weights: output/ssd_vgg16/model_final
+
+# Model Achitecture
+SSD:
+  # model feat info flow
+  backbone: VGG
+  ssd_head: SSDHead
+  # post process
+  post_process: BBoxPostProcess
+
+VGG:
+  depth: 16
+  normalizations: [20., -1, -1, -1, -1, -1]
+
+SSDHead:
+  in_channels: [512, 1024, 512, 256, 256, 256]
+  anchor_generator: AnchorGeneratorSSD
+
+AnchorGeneratorSSD:
+  steps: [8, 16, 32, 64, 100, 300]
+  aspect_ratios: [[2.], [2., 3.], [2., 3.], [2., 3.], [2.], [2.]]
+  min_ratio: 20
+  max_ratio: 90
+  min_sizes: [30.0, 60.0, 111.0, 162.0, 213.0, 264.0]
+  max_sizes: [60.0, 111.0, 162.0, 213.0, 264.0, 315.0]
+  offset: 0.5
+  flip: true
+  min_max_aspect_ratios_order: true
+
+BBoxPostProcess:
+  decode:
+    name: SSDBox
+  nms:
+    name: MultiClassNMS
+    keep_top_k: 200
+    score_threshold: 0.01
+    nms_threshold: 0.45
+    nms_top_k: 400
+    nms_eta: 1.0

configs/_base_/readers/ssd_reader.yml

+worker_num: 2
+TrainReader:
+  inputs_def:
+    fields: ['image', 'gt_bbox', 'gt_class']
+    num_max_boxes: 90
+
+  sample_transforms:
+    - DecodeOp: {}
+    - RandomDistortOp: {brightness: [0.5, 1.125, 0.875], random_apply: False}
+    - RandomExpandOp: {fill_value: [104., 117., 123.]}
+    - RandomCropOp: {allow_no_crop: true}
+    - RandomFlipOp: {}
+    - NormalizeBoxOp: {}
+    - ResizeOp: {target_size: [300, 300], keep_ratio: False, interp: 1}
+    - PadBoxOp: {num_max_boxes: 90}
+
+  batch_transforms:
+    - NormalizeImageOp: {mean: [104., 117., 123.], std: [1., 1., 1.], is_scale: false}
+    - PermuteOp: {}
+
+  batch_size: 8
+  shuffle: true
+  drop_last: true
+
+
+EvalReader:
+  inputs_def:
+    fields: ['image', 'im_shape',  'scale_factor', 'im_id', 'gt_bbox', 'gt_class', 'difficult']
+  sample_transforms:
+    - DecodeOp: {}
+    - ResizeOp: {target_size: [300, 300], keep_ratio: False, interp: 1}
+    - NormalizeImageOp: {mean: [104., 117., 123.], std: [1., 1., 1.], is_scale: false}
+    - PermuteOp: {}
+  batch_size: 1
+  drop_empty: false
+
+TestReader:
+  inputs_def:
+    image_shape: [3, 300, 300]
+    fields: ['image', 'im_shape', 'scale_factor', 'im_id']
+  sample_transforms:
+    - DecodeOp: {}
+    - ResizeOp: {target_size: [300, 300], keep_ratio: False, interp: 1}
+    - NormalizeImageOp: {mean: [104., 117., 123.], std: [1., 1., 1.], is_scale: false}
+    - PermuteOp: {}
+  batch_size: 1

configs/ssd_vgg16_300_120e_coco.yml

+_BASE_: [
+  './_base_/models/ssd_vgg16_300.yml',
+  './_base_/optimizers/ssd_120e.yml',
+  './_base_/datasets/coco.yml',
+  './_base_/readers/ssd_reader.yml',
+  './_base_/runtime.yml',
+]

configs/ssd_vgg16_300_240e_voc.yml

+_BASE_: [
+  './_base_/models/ssd_vgg16_300.yml',
+  './_base_/optimizers/ssd_240e.yml',
+  './_base_/datasets/voc.yml',
+  './_base_/readers/ssd_reader.yml',
+  './_base_/runtime.yml',
+]

deploy/python/infer.py

@@ -33,6 +33,7 @@ from paddle.inference import create_predictor
 SUPPORT_MODELS = {
     'YOLO',
     'RCNN',
+    'SSD',
 }
 
 
@@ -73,7 +74,7 @@ class Detector(object):
     def postprocess(self, np_boxes, np_masks, inputs, threshold=0.5):
         # postprocess output of predictor
         results = {}
-        if self.pred_config.arch in ['SSD', 'Face']:
+        if self.pred_config.arch in ['Face']:
             h, w = inputs['im_shape']
             scale_y, scale_x = inputs['scale_factor']
             w, h = float(h) / scale_y, float(w) / scale_x

docs/MODEL_ZOO_cn.md

@@ -41,3 +41,11 @@ Paddle提供基于ImageNet的骨架网络预训练模型。所有预训练模型
 | ResNet50-FPN         | Cascade Faster         |    1    |   1x    |     ----     |  41.1  |    -    | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/cascade_rcnn_r50_fpn_1x_coco.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/dygraph/configs/cascade_faster_rcnn_r50_fpn_1x_coco.yml) |
 | ResNet50-FPN         | Cascade Mask         |    1    |   1x    |     ----     |  41.6  |    35.3    | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/cascade_mask_rcnn_r50_fpn_1x_coco.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/dygraph/configs/cascade_mask_rcnn_r50_fpn_1x_coco.yml) |
 | DarkNet53         | YOLOv3         |    1    |   270e    |     ----     |  39.0  |    -    | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/yolov3_darknet53_270e_coco.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/dygraph/configs/yolov3_darknet53_270e_coco.yml) |
+
+### SSD on Pascal VOC
+
+| 骨架网络        | 网络类型       | 每张GPU图片个数 | 学习率策略 |推理时间(fps) | Box AP |                           下载                          | 配置文件 |
+| :-------------- | :------------- | :-----: | :-----: | :------------: | :-----: | :-----------------------------------------------------: | :-----: |
+| VGG             | SSD            |    8    |   240e    |     ----     |  78.2  | [下载链接](https://paddlemodels.bj.bcebos.com/object_detection/dygraph/ssd_vgg16_300_240e_voc.pdparams) | [配置文件](https://github.com/PaddlePaddle/PaddleDetection/tree/dygraph/configs/ssd_vgg16_300_240e_voc.yml) |
+
+**注意：** SSD使用4GPU训练，训练240个epoch

ppdet/data/reader.py

@@ -2,6 +2,7 @@ import copy
 import traceback
 import logging
 import threading
+import six
 import sys
 if sys.version_info >= (3, 0):
     import queue as Queue
@@ -118,25 +119,25 @@ class BaseDataLoader(object):
                  batch_sampler=None,
                  return_list=False,
                  use_prefetch=True):
-        self._dataset = dataset
-        self._dataset.parse_dataset(self.with_background)
+        self.dataset = dataset
+        self.dataset.parse_dataset(self.with_background)
         # get data
-        self._dataset.set_out(self._sample_transforms,
-                              copy.deepcopy(self._fields))
+        self.dataset.set_out(self._sample_transforms,
+                             copy.deepcopy(self._fields))
         # set kwargs
-        self._dataset.set_kwargs(**self.kwargs)
+        self.dataset.set_kwargs(**self.kwargs)
         # batch sampler
         if batch_sampler is None:
             self._batch_sampler = DistributedBatchSampler(
-                self._dataset,
+                self.dataset,
                 batch_size=self.batch_size,
                 shuffle=self.shuffle,
                 drop_last=self.drop_last)
         else:
             self._batch_sampler = batch_sampler
 
-        loader = DataLoader(
-            dataset=self._dataset,
+        self.loader = DataLoader(
+            dataset=self.dataset,
             batch_sampler=self._batch_sampler,
             collate_fn=self._batch_transforms,
             num_workers=worker_num,
@@ -144,8 +145,29 @@ class BaseDataLoader(object):
             return_list=return_list,
             use_buffer_reader=use_prefetch,
             use_shared_memory=False)
+        self.loader = iter(self.loader)
 
-        return loader, len(self._batch_sampler)
+        return self
+
+    def __len__(self):
+        return len(self._batch_sampler)
+
+    def __iter__(self):
+        return self
+
+    def __next__(self):
+        # pack {filed_name: field_data} here
+        # looking forward to support dictionary
+        # data structure in paddle.io.DataLoader
+        try:
+            data = next(self.loader)
+            return {k: v for k, v in zip(self._fields, data)}
+        except StopIteration:
+            six.reraise(*sys.exc_info())
+
+    def next(self):
+        # python2 compatibility
+        return self.__next__()
 
 
 @register

ppdet/data/source/__init__.py

@@ -14,9 +14,9 @@
 
 from . import coco
 # TODO add voc and widerface dataset
-#from . import voc
+from . import voc
 #from . import widerface
 
 from .coco import *
-#from .voc import *
+from .voc import *
 #from .widerface import *

ppdet/data/source/voc.py

@@ -19,14 +19,14 @@ import xml.etree.ElementTree as ET
 
 from ppdet.core.workspace import register, serializable
 
-from .dataset import DataSet
+from .dataset import DetDataset
 import logging
 logger = logging.getLogger(__name__)
 
 
 @register
 @serializable
-class VOCDataSet(DataSet):
+class VOCDataSet(DetDataset):
     """
     Load dataset with PascalVOC format.
 
@@ -38,8 +38,6 @@ class VOCDataSet(DataSet):
         image_dir (str): directory for images.
         anno_path (str): voc annotation file path.
         sample_num (int): number of samples to load, -1 means all.
-        use_default_label (bool): whether use the default mapping of
-            label to integer index. Default True.
         label_list (str): if use_default_label is False, will load
             mapping between category and class index.
     """
@@ -49,32 +47,15 @@ class VOCDataSet(DataSet):
                  image_dir=None,
                  anno_path=None,
                  sample_num=-1,
-                 use_default_label=True,
-                 label_list='label_list.txt'):
+                 label_list=None):
         super(VOCDataSet, self).__init__(
+            dataset_dir=dataset_dir,
             image_dir=image_dir,
             anno_path=anno_path,
-            sample_num=sample_num,
-            dataset_dir=dataset_dir)
-        # roidbs is list of dict whose structure is:
-        # {
-        #     'im_file': im_fname, # image file name
-        #     'im_id': im_id, # image id
-        #     'h': im_h, # height of image
-        #     'w': im_w, # width
-        #     'is_crowd': is_crowd,
-        #     'gt_class': gt_class,
-        #     'gt_score': gt_score,
-        #     'gt_bbox': gt_bbox,
-        #     'difficult': difficult
-        # }
-        self.roidbs = None
-        # 'cname2id' is a dict to map category name to class id
-        self.cname2cid = None
-        self.use_default_label = use_default_label
+            sample_num=sample_num)
         self.label_list = label_list
 
-    def load_roidb_and_cname2cid(self, with_background=True):
+    def parse_dataset(self, with_background=True):
         anno_path = os.path.join(self.dataset_dir, self.anno_path)
         image_dir = os.path.join(self.dataset_dir, self.image_dir)
 
@@ -86,7 +67,7 @@ class VOCDataSet(DataSet):
         records = []
         ct = 0
         cname2cid = {}
-        if not self.use_default_label:
+        if self.label_list:
             label_path = os.path.join(self.dataset_dir, self.label_list)
             if not os.path.exists(label_path):
                 raise ValueError("label_list {} does not exists".format(
@@ -183,6 +164,9 @@ class VOCDataSet(DataSet):
         logger.debug('{} samples in file {}'.format(ct, anno_path))
         self.roidbs, self.cname2cid = records, cname2cid
 
+    def get_label_list(self):
+        return os.path.join(self.dataset_dir, self.label_list)
+
 
 def pascalvoc_label(with_background=True):
     labels_map = {

ppdet/data/transform/operator.py

@@ -1668,7 +1668,7 @@ class PadBoxOp(BaseOperator):
         # in training, for example in op ExpandImage,
         # the bbox and gt_class is expandded, but the difficult is not,
         # so, judging by it's length
-        if 'is_difficult' in sample:
+        if 'difficult' in sample:
             pad_diff = np.zeros((num_max, ), dtype=np.int32)
             if gt_num > 0:
                 pad_diff[:gt_num] = sample['difficult'][:gt_num, 0]

ppdet/modeling/architecture/__init__.py

@@ -10,9 +10,11 @@ from . import faster_rcnn
 from . import mask_rcnn
 from . import yolo
 from . import cascade_rcnn
+from . import ssd
 
 from .meta_arch import *
 from .faster_rcnn import *
 from .mask_rcnn import *
 from .yolo import *
 from .cascade_rcnn import *
+from .ssd import *

ppdet/modeling/architecture/meta_arch.py

@@ -15,17 +15,8 @@ class BaseArch(nn.Layer):
     def __init__(self):
         super(BaseArch, self).__init__()
 
-    def forward(self,
-                input_tensor=None,
-                data=None,
-                input_def=None,
-                mode='infer'):
-        if input_tensor is None:
-            assert data is not None and input_def is not None
-            self.inputs = self.build_inputs(data, input_def)
-        else:
-            self.inputs = input_tensor
-
+    def forward(self, inputs, mode='infer'):
+        self.inputs = inputs
         self.inputs['mode'] = mode
         self.model_arch()
 

ppdet/modeling/architecture/ssd.py

+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from paddle import fluid
+from ppdet.core.workspace import register
+from .meta_arch import BaseArch
+
+__all__ = ['SSD']
+
+
+@register
+class SSD(BaseArch):
+    __category__ = 'architecture'
+    __inject__ = ['backbone', 'neck', 'ssd_head', 'post_process']
+
+    def __init__(self, backbone, ssd_head, post_process, neck=None):
+        super(SSD, self).__init__()
+        self.backbone = backbone
+        self.neck = neck
+        self.ssd_head = ssd_head
+        self.post_process = post_process
+
+    def model_arch(self):
+        # Backbone
+        body_feats = self.backbone(self.inputs)
+
+        # Neck
+        if self.neck is not None:
+            body_feats, spatial_scale = self.neck(body_feats)
+
+        # SSD Head
+        self.ssd_head_outs, self.anchors = self.ssd_head(body_feats,
+                                                         self.inputs['image'])
+
+    def get_loss(self, ):
+        loss = self.ssd_head.get_loss(self.ssd_head_outs, self.inputs,
+                                      self.anchors)
+        return {"loss": loss}
+
+    def get_pred(self, return_numpy=True):
+        output = {}
+        bbox, bbox_num = self.post_process(self.ssd_head_outs, self.anchors,
+                                           self.inputs['im_shape'],
+                                           self.inputs['scale_factor'])
+        outs = {
+            "bbox": bbox,
+            "bbox_num": bbox_num,
+        }
+        return outs

ppdet/modeling/backbone/__init__.py

+from . import vgg
 from . import resnet
 from . import darknet
 
+from .vgg import *
 from .resnet import *
 from .darknet import *

ppdet/modeling/backbone/vgg.py

+from __future__ import division
+
+import paddle
+import paddle.nn as nn
+import paddle.nn.functional as F
+from paddle import ParamAttr
+from paddle.fluid.regularizer import L2Decay
+from paddle.nn import Conv2D, MaxPool2D
+from ppdet.core.workspace import register, serializable
+
+__all__ = ['VGG']
+
+VGG_cfg = {16: [2, 2, 3, 3, 3], 19: [2, 2, 4, 4, 4]}
+
+
+class ConvBlock(nn.Layer):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 groups,
+                 pool_size=2,
+                 pool_stride=2,
+                 pool_padding=0,
+                 name=None):
+        super(ConvBlock, self).__init__()
+
+        self.groups = groups
+        self.conv0 = nn.Conv2D(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            weight_attr=ParamAttr(name=name + "1_weights"),
+            bias_attr=ParamAttr(name=name + "1_bias"))
+        self.conv_out_list = []
+        for i in range(1, groups):
+            conv_out = self.add_sublayer(
+                'conv{}'.format(i),
+                Conv2D(
+                    in_channels=out_channels,
+                    out_channels=out_channels,
+                    kernel_size=3,
+                    stride=1,
+                    padding=1,
+                    weight_attr=ParamAttr(
+                        name=name + "{}_weights".format(i + 1)),
+                    bias_attr=ParamAttr(name=name + "{}_bias".format(i + 1))))
+            self.conv_out_list.append(conv_out)
+
+        self.pool = MaxPool2D(
+            kernel_size=pool_size,
+            stride=pool_stride,
+            padding=pool_padding,
+            ceil_mode=True)
+
+    def forward(self, inputs):
+        out = self.conv0(inputs)
+        out = F.relu(out)
+        for conv_i in self.conv_out_list:
+            out = conv_i(out)
+            out = F.relu(out)
+        pool = self.pool(out)
+        return out, pool
+
+
+class ExtraBlock(nn.Layer):
+    def __init__(self,
+                 in_channels,
+                 mid_channels,
+                 out_channels,
+                 padding,
+                 stride,
+                 kernel_size,
+                 name=None):
+        super(ExtraBlock, self).__init__()
+
+        self.conv0 = Conv2D(
+            in_channels=in_channels,
+            out_channels=mid_channels,
+            kernel_size=1,
+            stride=1,
+            padding=0)
+        self.conv1 = Conv2D(
+            in_channels=mid_channels,
+            out_channels=out_channels,
+            kernel_size=kernel_size,
+            stride=stride,
+            padding=padding)
+
+    def forward(self, inputs):
+        out = self.conv0(inputs)
+        out = F.relu(out)
+        out = self.conv1(out)
+        out = F.relu(out)
+        return out
+
+
+class L2NormScale(nn.Layer):
+    def __init__(self, num_channels, scale=1.0):
+        super(L2NormScale, self).__init__()
+        self.scale = self.create_parameter(
+            attr=ParamAttr(initializer=paddle.nn.initializer.Constant(scale)),
+            shape=[num_channels])
+
+    def forward(self, inputs):
+        out = F.normalize(inputs, axis=1, epsilon=1e-10)
+        # out = self.scale.unsqueeze(0).unsqueeze(2).unsqueeze(3).expand_as(
+        #     out) * out
+        out = self.scale.unsqueeze(0).unsqueeze(2).unsqueeze(3) * out
+        return out
+
+
+@register
+@serializable
+class VGG(nn.Layer):
+    def __init__(self,
+                 depth=16,
+                 normalizations=[20., -1, -1, -1, -1, -1],
+                 extra_block_filters=[[256, 512, 1, 2, 3], [128, 256, 1, 2, 3],
+                                      [128, 256, 0, 1, 3],
+                                      [128, 256, 0, 1, 3]]):
+        super(VGG, self).__init__()
+
+        assert depth in [16, 19], \
+                "depth as 16/19 supported currently, but got {}".format(depth)
+        self.depth = depth
+        self.groups = VGG_cfg[depth]
+        self.normalizations = normalizations
+        self.extra_block_filters = extra_block_filters
+
+        self.conv_block_0 = ConvBlock(
+            3, 64, self.groups[0], 2, 2, 0, name="conv1_")
+        self.conv_block_1 = ConvBlock(
+            64, 128, self.groups[1], 2, 2, 0, name="conv2_")
+        self.conv_block_2 = ConvBlock(
+            128, 256, self.groups[2], 2, 2, 0, name="conv3_")
+        self.conv_block_3 = ConvBlock(
+            256, 512, self.groups[3], 2, 2, 0, name="conv4_")
+        self.conv_block_4 = ConvBlock(
+            512, 512, self.groups[4], 3, 1, 1, name="conv5_")
+
+        self.fc6 = Conv2D(
+            in_channels=512,
+            out_channels=1024,
+            kernel_size=3,
+            stride=1,
+            padding=6,
+            dilation=6)
+        self.fc7 = Conv2D(
+            in_channels=1024,
+            out_channels=1024,
+            kernel_size=1,
+            stride=1,
+            padding=0)
+
+        # extra block
+        self.extra_convs = []
+        last_channels = 1024
+        for i, v in enumerate(self.extra_block_filters):
+            assert len(v) == 5, "extra_block_filters size not fix"
+            extra_conv = self.add_sublayer("conv{}".format(6 + i),
+                                           ExtraBlock(last_channels, v[0], v[1],
+                                                      v[2], v[3], v[4]))
+            last_channels = v[1]
+            self.extra_convs.append(extra_conv)
+
+        self.norms = []
+        for i, n in enumerate(self.normalizations):
+            if n != -1:
+                norm = self.add_sublayer("norm{}".format(i),
+                                         L2NormScale(
+                                             self.extra_block_filters[i][1], n))
+            else:
+                norm = None
+            self.norms.append(norm)
+
+    def forward(self, inputs):
+        outputs = []
+
+        conv, pool = self.conv_block_0(inputs['image'])
+        conv, pool = self.conv_block_1(pool)
+        conv, pool = self.conv_block_2(pool)
+        conv, pool = self.conv_block_3(pool)
+        outputs.append(conv)
+
+        conv, pool = self.conv_block_4(pool)
+        out = self.fc6(pool)
+        out = F.relu(out)
+        out = self.fc7(out)
+        out = F.relu(out)
+        outputs.append(out)
+
+        if not self.extra_block_filters:
+            return out
+
+        # extra block
+        for extra_conv in self.extra_convs:
+            out = extra_conv(out)
+            outputs.append(out)
+
+        for i, n in enumerate(self.normalizations):
+            if n != -1:
+                outputs[i] = self.norms[i](outputs[i])
+
+        return outputs

ppdet/modeling/head/__init__.py

@@ -17,9 +17,11 @@ from . import bbox_head
 from . import mask_head
 from . import yolo_head
 from . import roi_extractor
+from . import ssd_head
 
 from .rpn_head import *
 from .bbox_head import *
 from .mask_head import *
 from .yolo_head import *
 from .roi_extractor import *
+from .ssd_head import *

ppdet/modeling/head/ssd_head.py

+import paddle
+import paddle.nn as nn
+import paddle.nn.functional as F
+import numpy as np
+from ppdet.core.workspace import register
+
+
+@register
+class SSDHead(nn.Layer):
+    __shared__ = ['num_classes']
+    __inject__ = ['anchor_generator', 'loss']
+
+    def __init__(self,
+                 num_classes=81,
+                 in_channels=(512, 1024, 512, 256, 256, 256),
+                 anchor_generator='AnchorGeneratorSSD',
+                 loss='SSDLoss'):
+        super(SSDHead, self).__init__()
+        self.num_classes = num_classes
+        self.in_channels = in_channels
+        self.anchor_generator = anchor_generator
+        self.loss = loss
+        self.num_priors = self.anchor_generator.num_priors
+
+        self.box_convs = []
+        self.score_convs = []
+        for i, num_prior in enumerate(self.num_priors):
+            self.box_convs.append(
+                self.add_sublayer(
+                    "boxes{}".format(i),
+                    nn.Conv2D(
+                        in_channels=in_channels[i],
+                        out_channels=num_prior * 4,
+                        kernel_size=3,
+                        padding=1)))
+            self.score_convs.append(
+                self.add_sublayer(
+                    "scores{}".format(i),
+                    nn.Conv2D(
+                        in_channels=in_channels[i],
+                        out_channels=num_prior * num_classes,
+                        kernel_size=3,
+                        padding=1)))
+
+    def forward(self, feats, image):
+        box_preds = []
+        cls_scores = []
+        prior_boxes = []
+        for feat, box_conv, score_conv in zip(feats, self.box_convs,
+                                              self.score_convs):
+            box_pred = box_conv(feat)
+            box_pred = paddle.transpose(box_pred, [0, 2, 3, 1])
+            box_pred = paddle.reshape(box_pred, [0, -1, 4])
+            box_preds.append(box_pred)
+
+            cls_score = score_conv(feat)
+            cls_score = paddle.transpose(cls_score, [0, 2, 3, 1])
+            cls_score = paddle.reshape(cls_score, [0, -1, self.num_classes])
+            cls_scores.append(cls_score)
+
+        prior_boxes = self.anchor_generator(feats, image)
+
+        outputs = {}
+        outputs['boxes'] = box_preds
+        outputs['scores'] = cls_scores
+        return outputs, prior_boxes
+
+    def get_loss(self, inputs, targets, prior_boxes):
+        return self.loss(inputs, targets, prior_boxes)

ppdet/modeling/layers.py

@@ -12,6 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import math
 import numpy as np
 from numbers import Integral
 
@@ -24,6 +25,12 @@ from . import ops
 import paddle.nn.functional as F
 
 
+def _to_list(l):
+    if isinstance(l, (list, tuple)):
+        return list(l)
+    return [l]
+
+
 @register
 @serializable
 class AnchorGeneratorRPN(object):
@@ -103,6 +110,57 @@ class AnchorTargetGeneratorRPN(object):
         return pred_cls_logits, pred_bbox_pred, tgt_labels, tgt_bboxes, bbox_inside_weights
 
 
+@register
+@serializable
+class AnchorGeneratorSSD(object):
+    def __init__(self,
+                 steps=[8, 16, 32, 64, 100, 300],
+                 aspect_ratios=[[2.], [2., 3.], [2., 3.], [2., 3.], [2.], [2.]],
+                 min_ratio=15,
+                 max_ratio=90,
+                 min_sizes=[30.0, 60.0, 111.0, 162.0, 213.0, 264.0],
+                 max_sizes=[60.0, 111.0, 162.0, 213.0, 264.0, 315.0],
+                 offset=0.5,
+                 flip=True,
+                 clip=False,
+                 min_max_aspect_ratios_order=False):
+        self.steps = steps
+        self.aspect_ratios = aspect_ratios
+        self.min_ratio = min_ratio
+        self.max_ratio = max_ratio
+        self.min_sizes = min_sizes
+        self.max_sizes = max_sizes
+        self.offset = offset
+        self.flip = flip
+        self.clip = clip
+        self.min_max_aspect_ratios_order = min_max_aspect_ratios_order
+
+        self.num_priors = []
+        for aspect_ratio, min_size, max_size in zip(aspect_ratios, min_sizes,
+                                                    max_sizes):
+            self.num_priors.append((len(aspect_ratio) * 2 + 1) * len(
+                _to_list(min_size)) + len(_to_list(max_size)))
+
+    def __call__(self, inputs, image):
+        boxes = []
+        for input, min_size, max_size, aspect_ratio, step in zip(
+                inputs, self.min_sizes, self.max_sizes, self.aspect_ratios,
+                self.steps):
+            box, _ = ops.prior_box(
+                input=input,
+                image=image,
+                min_sizes=_to_list(min_size),
+                max_sizes=_to_list(max_size),
+                aspect_ratios=aspect_ratio,
+                flip=self.flip,
+                clip=self.clip,
+                steps=[step, step],
+                offset=self.offset,
+                min_max_aspect_ratios_order=self.min_max_aspect_ratios_order)
+            boxes.append(paddle.reshape(box, [-1, 4]))
+        return boxes
+
+
 @register
 @serializable
 class ProposalGenerator(object):
@@ -420,7 +478,12 @@ class YOLOBox(object):
         self.clip_bbox = clip_bbox
         self.scale_x_y = scale_x_y
 
-    def __call__(self, yolo_head_out, anchors, im_shape, scale_factor):
+    def __call__(self,
+                 yolo_head_out,
+                 anchors,
+                 im_shape,
+                 scale_factor,
+                 var_weight=None):
         boxes_list = []
         scores_list = []
         origin_shape = im_shape / scale_factor
@@ -437,6 +500,54 @@ class YOLOBox(object):
         return yolo_boxes, yolo_scores
 
 
+@register
+@serializable
+class SSDBox(object):
+    def __init__(self, is_normalized=True):
+        self.is_normalized = is_normalized
+        self.norm_delta = float(not self.is_normalized)
+
+    def __call__(self,
+                 preds,
+                 prior_boxes,
+                 im_shape,
+                 scale_factor,
+                 var_weight=None):
+        boxes, scores = preds['boxes'], preds['scores']
+        outputs = []
+        for box, score, prior_box in zip(boxes, scores, prior_boxes):
+            pb_w = prior_box[:, 2] - prior_box[:, 0] + self.norm_delta
+            pb_h = prior_box[:, 3] - prior_box[:, 1] + self.norm_delta
+            pb_x = prior_box[:, 0] + pb_w * 0.5
+            pb_y = prior_box[:, 1] + pb_h * 0.5
+            out_x = pb_x + box[:, :, 0] * pb_w * 0.1
+            out_y = pb_y + box[:, :, 1] * pb_h * 0.1
+            out_w = paddle.exp(box[:, :, 2] * 0.2) * pb_w
+            out_h = paddle.exp(box[:, :, 3] * 0.2) * pb_h
+
+            if self.is_normalized:
+                h = im_shape[:, 0] / scale_factor[:, 0]
+                w = im_shape[:, 1] / scale_factor[:, 1]
+                output = paddle.stack(
+                    [(out_x - out_w / 2.) * w, (out_y - out_h / 2.) * h,
+                     (out_x + out_w / 2.) * w, (out_y + out_h / 2.) * h],
+                    axis=-1)
+            else:
+                output = paddle.stack(
+                    [
+                        out_x - out_w / 2., out_y - out_h / 2.,
+                        out_x + out_w / 2. - 1., out_y + out_h / 2. - 1.
+                    ],
+                    axis=-1)
+            outputs.append(output)
+        boxes = paddle.concat(outputs, axis=1)
+
+        scores = F.softmax(paddle.concat(scores, axis=1))
+        scores = paddle.transpose(scores, [0, 2, 1])
+
+        return boxes, scores
+
+
 @register
 @serializable
 class AnchorGrid(object):

ppdet/modeling/loss/__init__.py

@@ -15,7 +15,9 @@
 from . import yolo_loss
 from . import iou_aware_loss
 from . import iou_loss
+from . import ssd_loss
 
 from .yolo_loss import *
 from .iou_aware_loss import *
 from .iou_loss import *
+from .ssd_loss import *

ppdet/modeling/loss/ssd_loss.py

+# 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.
+
+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
+import numpy as np
+from ppdet.core.workspace import register
+from ..ops import bipartite_match, box_coder, iou_similarity
+
+__all__ = ['SSDLoss']
+
+
+@register
+class SSDLoss(nn.Layer):
+    def __init__(self,
+                 match_type='per_prediction',
+                 overlap_threshold=0.5,
+                 neg_pos_ratio=3.0,
+                 neg_overlap=0.5,
+                 loc_loss_weight=1.0,
+                 conf_loss_weight=1.0):
+        super(SSDLoss, self).__init__()
+        self.match_type = match_type
+        self.overlap_threshold = overlap_threshold
+        self.neg_pos_ratio = neg_pos_ratio
+        self.neg_overlap = neg_overlap
+        self.loc_loss_weight = loc_loss_weight
+        self.conf_loss_weight = conf_loss_weight
+
+    def _label_target_assign(self,
+                             gt_label,
+                             matched_indices,
+                             neg_mask=None,
+                             mismatch_value=0):
+        gt_label = gt_label.numpy()
+        matched_indices = matched_indices.numpy()
+        if neg_mask is not None:
+            neg_mask = neg_mask.numpy()
+
+        batch_size, num_priors = matched_indices.shape
+        trg_lbl = np.ones((batch_size, num_priors, 1)).astype('int32')
+        trg_lbl *= mismatch_value
+        trg_lbl_wt = np.zeros((batch_size, num_priors, 1)).astype('float32')
+
+        for i in range(batch_size):
+            col_ids = np.where(matched_indices[i] > -1)
+            col_val = matched_indices[i][col_ids]
+            trg_lbl[i][col_ids] = gt_label[i][col_val]
+            trg_lbl_wt[i][col_ids] = 1.0
+
+        if neg_mask is not None:
+            trg_lbl_wt += neg_mask[:, :, np.newaxis]
+
+        return paddle.to_tensor(trg_lbl), paddle.to_tensor(trg_lbl_wt)
+
+    def _bbox_target_assign(self, encoded_box, matched_indices):
+        encoded_box = encoded_box.numpy()
+        matched_indices = matched_indices.numpy()
+
+        batch_size, num_priors = matched_indices.shape
+        trg_bbox = np.zeros((batch_size, num_priors, 4)).astype('float32')
+        trg_bbox_wt = np.zeros((batch_size, num_priors, 1)).astype('float32')
+
+        for i in range(batch_size):
+            col_ids = np.where(matched_indices[i] > -1)
+            col_val = matched_indices[i][col_ids]
+            for v, c in zip(col_val.tolist(), col_ids[0]):
+                trg_bbox[i][c] = encoded_box[i][v][c]
+            trg_bbox_wt[i][col_ids] = 1.0
+
+        return paddle.to_tensor(trg_bbox), paddle.to_tensor(trg_bbox_wt)
+
+    def _mine_hard_example(self,
+                           conf_loss,
+                           matched_indices,
+                           matched_dist,
+                           neg_pos_ratio=3.0,
+                           neg_overlap=0.5):
+        pos = (matched_indices > -1).astype(conf_loss.dtype)
+        num_pos = pos.sum(axis=1, keepdim=True)
+        neg = (matched_dist < neg_overlap).astype(conf_loss.dtype)
+
+        conf_loss = conf_loss * (1.0 - pos) * neg
+        loss_idx = conf_loss.argsort(axis=1, descending=True)
+        idx_rank = loss_idx.argsort(axis=1)
+        num_negs = []
+        for i in range(matched_indices.shape[0]):
+            cur_idx = loss_idx[i]
+            cur_num_pos = num_pos[i]
+            num_neg = paddle.clip(cur_num_pos * neg_pos_ratio, max=pos.shape[1])
+            num_negs.append(num_neg)
+        num_neg = paddle.stack(num_negs, axis=0).expand_as(idx_rank)
+        neg_mask = (idx_rank < num_neg).astype(conf_loss.dtype)
+        return neg_mask
+
+    def forward(self, inputs, targets, anchors):
+        boxes = paddle.concat(inputs['boxes'], axis=1)
+        scores = paddle.concat(inputs['scores'], axis=1)
+        prior_boxes = paddle.concat(anchors, axis=0)
+        gt_box = targets['gt_bbox']
+        gt_label = targets['gt_class'].unsqueeze(-1)
+        batch_size, num_priors, num_classes = scores.shape
+
+        def _reshape_to_2d(x):
+            return paddle.flatten(x, start_axis=2)
+
+        # 1. Find matched bounding box by prior box.
+        #   1.1 Compute IOU similarity between ground-truth boxes and prior boxes.
+        #   1.2 Compute matched bounding box by bipartite matching algorithm.
+        matched_indices = []
+        matched_dist = []
+        for i in range(gt_box.shape[0]):
+            iou = iou_similarity(gt_box[i], prior_boxes)
+            matched_indice, matched_d = bipartite_match(iou, self.match_type,
+                                                        self.overlap_threshold)
+            matched_indices.append(matched_indice)
+            matched_dist.append(matched_d)
+        matched_indices = paddle.concat(matched_indices, axis=0)
+        matched_indices.stop_gradient = True
+        matched_dist = paddle.concat(matched_dist, axis=0)
+        matched_dist.stop_gradient = True
+
+        # 2. Compute confidence for mining hard examples
+        # 2.1. Get the target label based on matched indices
+        target_label, _ = self._label_target_assign(gt_label, matched_indices)
+        confidence = _reshape_to_2d(scores)
+        # 2.2. Compute confidence loss.
+        # Reshape confidence to 2D tensor.
+        target_label = _reshape_to_2d(target_label).astype('int64')
+        conf_loss = F.softmax_with_cross_entropy(confidence, target_label)
+        conf_loss = paddle.reshape(conf_loss, [batch_size, num_priors])
+
+        # 3. Mining hard examples
+        neg_mask = self._mine_hard_example(
+            conf_loss,
+            matched_indices,
+            matched_dist,
+            neg_pos_ratio=self.neg_pos_ratio,
+            neg_overlap=self.neg_overlap)
+
+        # 4. Assign classification and regression targets
+        # 4.1. Encoded bbox according to the prior boxes.
+        prior_box_var = paddle.to_tensor(
+            np.array(
+                [0.1, 0.1, 0.2, 0.2], dtype='float32')).reshape(
+                    [1, 4]).expand_as(prior_boxes)
+        encoded_bbox = []
+        for i in range(gt_box.shape[0]):
+            encoded_bbox.append(
+                box_coder(
+                    prior_box=prior_boxes,
+                    prior_box_var=prior_box_var,
+                    target_box=gt_box[i],
+                    code_type='encode_center_size'))
+        encoded_bbox = paddle.stack(encoded_bbox, axis=0)
+        # 4.2. Assign regression targets
+        target_bbox, target_loc_weight = self._bbox_target_assign(
+            encoded_bbox, matched_indices)
+        # 4.3. Assign classification targets
+        target_label, target_conf_weight = self._label_target_assign(
+            gt_label, matched_indices, neg_mask=neg_mask)
+
+        # 5. Compute loss.
+        # 5.1 Compute confidence loss.
+        target_label = _reshape_to_2d(target_label).astype('int64')
+        conf_loss = F.softmax_with_cross_entropy(confidence, target_label)
+
+        target_conf_weight = _reshape_to_2d(target_conf_weight)
+        conf_loss = conf_loss * target_conf_weight * self.conf_loss_weight
+
+        # 5.2 Compute regression loss.
+        location = _reshape_to_2d(boxes)
+        target_bbox = _reshape_to_2d(target_bbox)
+
+        loc_loss = F.smooth_l1_loss(location, target_bbox, reduction='none')
+        loc_loss = paddle.sum(loc_loss, axis=-1, keepdim=True)
+        target_loc_weight = _reshape_to_2d(target_loc_weight)
+        loc_loss = loc_loss * target_loc_weight * self.loc_loss_weight
+
+        # 5.3 Compute overall weighted loss.
+        loss = conf_loss + loc_loss
+        loss = paddle.reshape(loss, [batch_size, num_priors])
+        loss = paddle.sum(loss, axis=1, keepdim=True)
+        normalizer = paddle.sum(target_loc_weight)
+        loss = paddle.sum(loss / normalizer)
+
+        return loss

ppdet/modeling/ops.py

@@ -806,15 +806,12 @@ def prior_box(input,
         cur_max_sizes = max_sizes
 
     if in_dygraph_mode():
-        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.extend('max_sizes', max_sizes)
-        attrs = tuple(attrs)
+        assert cur_max_sizes is not None
+        attrs = ('min_sizes', min_sizes, 'max_sizes', cur_max_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)
         box, var = core.ops.prior_box(input, image, *attrs)
         return box, var
     else:
@@ -1254,6 +1251,111 @@ def matrix_nms(bboxes,
         return output
 
 
+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,

ppdet/modeling/tests/test_ops.py

@@ -376,6 +376,31 @@ class TestIoUSimilarity(LayerTest):
         self.assertTrue(np.array_equal(iou_np, iou_dy_np))
 
 
+class TestBipartiteMatch(LayerTest):
+    def test_bipartite_match(self):
+        distance = np.random.random((20, 10)).astype('float32')
+        with self.static_graph():
+            x = paddle.static.data(name='x', shape=[20, 10], dtype='float32')
+
+            match_indices, match_dist = ops.bipartite_match(
+                x, match_type='per_prediction', dist_threshold=0.5)
+            match_indices_np, match_dist_np = self.get_static_graph_result(
+                feed={'x': distance, },
+                fetch_list=[match_indices, match_dist],
+                with_lod=False)
+
+        with self.dynamic_graph():
+            x_dy = base.to_variable(distance)
+
+            match_indices_dy, match_dist_dy = ops.bipartite_match(
+                x_dy, match_type='per_prediction', dist_threshold=0.5)
+            match_indices_dy_np = match_indices_dy.numpy()
+            match_dist_dy_np = match_dist_dy.numpy()
+
+        self.assertTrue(np.array_equal(match_indices_np, match_indices_dy_np))
+        self.assertTrue(np.array_equal(match_dist_np, match_dist_dy_np))
+
+
 class TestYoloBox(LayerTest):
     def test_yolo_box(self):
 

ppdet/utils/eval_utils.py

@@ -33,7 +33,7 @@ def json_eval_results(metric, json_directory=None, dataset=None):
             logger.info("{} not exists!".format(v_json))
 
 
-def get_infer_results(outs_res, eval_type, catid, im_info):
+def get_infer_results(outs_res, eval_type, catid):
     """
     Get result at the stage of inference.
     The output format is dictionary containing bbox or mask result.
@@ -45,31 +45,27 @@ def get_infer_results(outs_res, eval_type, catid, im_info):
         raise ValueError(
             'The number of valid detection result if zero. Please use reasonable model and check input data.'
         )
-    infer_res = {}
-
-    if 'bbox' in eval_type:
-        box_res = []
-        for i, outs in enumerate(outs_res):
-            im_ids = im_info[i][2]
-            box_res += get_det_res(outs['bbox'], outs['bbox_num'], im_ids,
-                                   catid)
-        infer_res['bbox'] = box_res
-
-    if 'mask' in eval_type:
-        seg_res = []
-        # mask post process
-        for i, outs in enumerate(outs_res):
-            im_shape = im_info[i][0]
-            scale_factor = im_info[i][1]
-            im_ids = im_info[i][2]
-            mask = outs['mask']
-            seg_res += get_seg_res(mask, outs['bbox_num'], im_ids, catid)
-        infer_res['mask'] = seg_res
+
+    infer_res = {k: [] for k in eval_type}
+
+    for i, outs in enumerate(outs_res):
+        im_id = outs['im_id']
+        im_shape = outs['im_shape']
+        scale_factor = outs['scale_factor']
+
+        if 'bbox' in eval_type:
+            infer_res['bbox'] += get_det_res(outs['bbox'], outs['bbox_num'],
+                                             im_id, catid)
+
+        if 'mask' in eval_type:
+            # mask post process
+            infer_res['mask'] += get_seg_res(outs['mask'], outs['bbox_num'],
+                                             im_id, catid)
 
     return infer_res
 
 
-def eval_results(res, metric, anno_file):
+def eval_results(res, metric, dataset):
     """
     Evalute the inference result
     """
@@ -82,7 +78,8 @@ def eval_results(res, metric, anno_file):
                 json.dump(res['bbox'], f)
                 logger.info('The bbox result is saved to bbox.json.')
 
-            bbox_stats = cocoapi_eval('bbox.json', 'bbox', anno_file=anno_file)
+            bbox_stats = cocoapi_eval(
+                'bbox.json', 'bbox', anno_file=dataset.get_anno())
             eval_res.append(bbox_stats)
             sys.stdout.flush()
         if 'mask' in res:
@@ -90,9 +87,14 @@ def eval_results(res, metric, anno_file):
                 json.dump(res['mask'], f)
                 logger.info('The mask result is saved to mask.json.')
 
-            seg_stats = cocoapi_eval('mask.json', 'segm', anno_file=anno_file)
+            seg_stats = cocoapi_eval(
+                'mask.json', 'segm', anno_file=dataset.get_anno())
             eval_res.append(seg_stats)
             sys.stdout.flush()
+    elif metric == 'VOC':
+        from ppdet.utils.voc_eval import bbox_eval
+
+        bbox_stats = bbox_eval(res, 21)
     else:
         raise NotImplemented("Only COCO metric is supported now.")
 

ppdet/utils/voc_eval.py

@@ -63,46 +63,33 @@ def bbox_eval(results,
         evaluate_difficult=evaluate_difficult)
 
     for t in results:
-        bboxes = t['bbox'][0]
-        bbox_lengths = t['bbox'][1][0]
+        bboxes = t['bbox']
+        bbox_lengths = t['bbox_num']
 
         if bboxes.shape == (1, 1) or bboxes is None:
             continue
-        gt_boxes = t['gt_bbox'][0]
-        gt_labels = t['gt_class'][0]
-        difficults = t['is_difficult'][0] if not evaluate_difficult \
+        gt_boxes = t['gt_bbox']
+        gt_labels = t['gt_class']
+        difficults = t['difficult'] if not evaluate_difficult \
                             else None
 
-        if len(t['gt_bbox'][1]) == 0:
-            # gt_bbox, gt_class, difficult read as zero padded Tensor
-            bbox_idx = 0
-            for i in range(len(gt_boxes)):
-                gt_box = gt_boxes[i]
-                gt_label = gt_labels[i]
-                difficult = None if difficults is None \
-                                else difficults[i]
-                bbox_num = bbox_lengths[i]
-                bbox = bboxes[bbox_idx:bbox_idx + bbox_num]
-                gt_box, gt_label, difficult = prune_zero_padding(
-                    gt_box, gt_label, difficult)
-                detection_map.update(bbox, gt_box, gt_label, difficult)
-                bbox_idx += bbox_num
-        else:
-            # gt_box, gt_label, difficult read as LoDTensor
-            gt_box_lengths = t['gt_bbox'][1][0]
-            bbox_idx = 0
-            gt_box_idx = 0
-            for i in range(len(bbox_lengths)):
-                bbox_num = bbox_lengths[i]
-                gt_box_num = gt_box_lengths[i]
-                bbox = bboxes[bbox_idx:bbox_idx + bbox_num]
-                gt_box = gt_boxes[gt_box_idx:gt_box_idx + gt_box_num]
-                gt_label = gt_labels[gt_box_idx:gt_box_idx + gt_box_num]
-                difficult = None if difficults is None else \
-                            difficults[gt_box_idx: gt_box_idx + gt_box_num]
-                detection_map.update(bbox, gt_box, gt_label, difficult)
-                bbox_idx += bbox_num
-                gt_box_idx += gt_box_num
+        scale_factor = t['scale_factor'] if 'scale_factor' in t else np.ones(
+            (gt_boxes.shape[0], 2)).astype('float32')
+
+        bbox_idx = 0
+        for i in range(gt_boxes.shape[0]):
+            gt_box = gt_boxes[i]
+            h, w = scale_factor[i]
+            gt_box = gt_box / np.array([w, h, w, h])
+            gt_label = gt_labels[i]
+            difficult = None if difficults is None \
+                            else difficults[i]
+            bbox_num = bbox_lengths[i]
+            bbox = bboxes[bbox_idx:bbox_idx + bbox_num]
+            gt_box, gt_label, difficult = prune_zero_padding(gt_box, gt_label,
+                                                             difficult)
+            detection_map.update(bbox, gt_box, gt_label, difficult)
+            bbox_idx += bbox_num
 
     logger.info("Accumulating evaluatation results...")
     detection_map.accumulate()

tools/eval.py

@@ -69,34 +69,34 @@ def run(FLAGS, cfg, place):
 
     # Data Reader
     dataset = cfg.EvalDataset
-    eval_loader, _ = create('EvalReader')(dataset, cfg['worker_num'], place)
+    eval_loader = create('EvalReader')(dataset, cfg['worker_num'], place)
+
+    extra_key = ['im_shape', 'scale_factor', 'im_id']
+    if cfg.metric == 'VOC':
+        extra_key += ['gt_bbox', 'gt_class', 'difficult']
 
     # Run Eval
     outs_res = []
     start_time = time.time()
     sample_num = 0
-    im_info = []
     for iter_id, data in enumerate(eval_loader):
         # forward
-        fields = cfg['EvalReader']['inputs_def']['fields']
         model.eval()
-        outs = model(data=data, input_def=fields, mode='infer')
+        outs = model(data, mode='infer')
+        for key in extra_key:
+            outs[key] = data[key]
         for key, value in outs.items():
             outs[key] = value.numpy()
-        im_shape = data[fields.index('im_shape')].numpy()
-        scale_factor = data[fields.index('scale_factor')].numpy()
-        im_id = data[fields.index('im_id')].numpy()
-        im_info.append([im_shape, scale_factor, im_id])
 
         if 'mask' in outs and 'bbox' in outs:
             mask_resolution = model.mask_post_process.mask_resolution
             from ppdet.py_op.post_process import mask_post_process
-            outs['mask'] = mask_post_process(outs, im_shape, scale_factor,
-                                             mask_resolution)
+            outs['mask'] = mask_post_process(
+                outs, outs['im_shape'], outs['scale_factor'], mask_resolution)
 
         outs_res.append(outs)
         # log
-        sample_num += im_shape.shape[0]
+        sample_num += outs['im_id'].shape[0]
         if iter_id % 100 == 0:
             logger.info("Eval iter: {}".format(iter_id))
 
@@ -111,15 +111,22 @@ def run(FLAGS, cfg, place):
         eval_type.append('mask')
     # Metric
     # TODO: support other metric
-    from ppdet.utils.coco_eval import get_category_info
-    anno_file = dataset.get_anno()
     with_background = cfg.with_background
     use_default_label = dataset.use_default_label
-    clsid2catid, catid2name = get_category_info(anno_file, with_background,
-                                                use_default_label)
+    if cfg.metric == 'COCO':
+        from ppdet.utils.coco_eval import get_category_info
+        clsid2catid, catid2name = get_category_info(
+            dataset.get_anno(), with_background, use_default_label)
+
+        infer_res = get_infer_results(outs_res, eval_type, clsid2catid)
+
+    elif cfg.metric == 'VOC':
+        from ppdet.utils.voc_eval import get_category_info
+        clsid2catid, catid2name = get_category_info(
+            dataset.get_label_list(), with_background, use_default_label)
+        infer_res = outs_res
 
-    infer_res = get_infer_results(outs_res, eval_type, clsid2catid, im_info)
-    eval_results(infer_res, cfg.metric, anno_file)
+    eval_results(infer_res, cfg.metric, dataset)
 
 
 def main():

tools/export_utils.py

@@ -49,6 +49,8 @@ def parse_reader(reader_cfg, dataset_cfg, metric, arch, image_shape):
 
     if metric == 'COCO':
         from ppdet.utils.coco_eval import get_category_info
+    elif metric == 'VOC':
+        from ppdet.utils.voc_eval import get_category_info
     else:
         raise ValueError("metric only supports COCO, but received {}".format(
             metric))

tools/infer.py

@@ -129,7 +129,11 @@ def run(FLAGS, cfg, place):
     dataset = cfg.TestDataset
     test_images = get_test_images(FLAGS.infer_dir, FLAGS.infer_img)
     dataset.set_images(test_images)
-    test_loader, _ = create('TestReader')(dataset, cfg['worker_num'], place)
+    test_loader = create('TestReader')(dataset, cfg['worker_num'], place)
+
+    extra_key = ['im_shape', 'scale_factor', 'im_id']
+    if cfg.metric == 'VOC':
+        extra_key += ['gt_bbox', 'gt_class', 'difficult']
 
     # TODO: support other metrics
     imid2path = dataset.get_imid2path()
@@ -147,24 +151,18 @@ def run(FLAGS, cfg, place):
     # Run Infer 
     for iter_id, data in enumerate(test_loader):
         # forward
-        fields = cfg.TestReader['inputs_def']['fields']
         model.eval()
-        outs = model(
-            data=data,
-            input_def=cfg.TestReader['inputs_def']['fields'],
-            mode='infer')
+        outs = model(data, mode='infer')
+        for key in extra_key:
+            outs[key] = data[key]
         for key, value in outs.items():
             outs[key] = value.numpy()
-        im_shape = data[fields.index('im_shape')].numpy()
-        scale_factor = data[fields.index('scale_factor')].numpy()
-        im_ids = data[fields.index('im_id')].numpy()
-        im_info = [im_shape, scale_factor, im_ids]
 
         if 'mask' in outs and 'bbox' in outs:
             mask_resolution = model.mask_post_process.mask_resolution
             from ppdet.py_op.post_process import mask_post_process
-            outs['mask'] = mask_post_process(outs, im_shape, scale_factor,
-                                             mask_resolution)
+            outs['mask'] = mask_post_process(
+                outs, outs['im_shape'], outs['scale_factor'], mask_resolution)
 
         eval_type = []
         if 'bbox' in outs:
@@ -172,14 +170,14 @@ def run(FLAGS, cfg, place):
         if 'mask' in outs:
             eval_type.append('mask')
 
-        batch_res = get_infer_results([outs], eval_type, clsid2catid, [im_info])
+        batch_res = get_infer_results([outs], eval_type, clsid2catid)
         logger.info('Infer iter {}'.format(iter_id))
         bbox_res = None
         mask_res = None
 
         bbox_num = outs['bbox_num']
         start = 0
-        for i, im_id in enumerate(im_ids):
+        for i, im_id in enumerate(outs['im_id']):
             image_path = imid2path[int(im_id)]
             image = Image.open(image_path).convert('RGB')
             end = start + bbox_num[i]
@@ -197,7 +195,7 @@ def run(FLAGS, cfg, place):
                 mask_res = batch_res['mask'][start:end]
 
             image = visualize_results(image, bbox_res, mask_res,
-                                      int(im_id), catid2name,
+                                      int(outs['im_id']), catid2name,
                                       FLAGS.draw_threshold)
 
             # use VisualDL to log image with bbox

tools/train.py

@@ -103,8 +103,8 @@ def run(FLAGS, cfg, place):
 
     # Data 
     dataset = cfg.TrainDataset
-    train_loader, step_per_epoch = create('TrainReader')(
-        dataset, cfg['worker_num'], place)
+    train_loader = create('TrainReader')(dataset, cfg['worker_num'], place)
+    step_per_epoch = len(train_loader)
 
     # Model
     model = create(cfg.architecture)
@@ -134,7 +134,6 @@ def run(FLAGS, cfg, place):
     if ParallelEnv().nranks > 1:
         model = paddle.DataParallel(model)
 
-    fields = train_loader.collate_fn.output_fields
     cfg_name = os.path.basename(FLAGS.config).split('.')[0]
     save_dir = os.path.join(cfg.save_dir, cfg_name)
     # Run Train
@@ -155,7 +154,7 @@ def run(FLAGS, cfg, place):
 
             # Model Forward
             model.train()
-            outputs = model(data=data, input_def=fields, mode='train')
+            outputs = model(data, mode='train')
 
             # Model Backward
             loss = outputs['loss']