Merge pull request #2184 from open-mmlab/dev-1.x

Dev 1.x

README.md

@@ -130,36 +130,36 @@ MMPose v1.0.0 is a major update, including many API and config file changes. Cur
 <details close>
 <summary><b>Migration Progress</b></summary>
 
-| Algorithm | Status |
-| :--- | :---: |
-| MTUT (CVPR 2019) |  | 
-| MSPN (ArXiv 2019)| done | 
-| InterNet (ECCV 2020) |  | 
-| DEKR (CVPR 2021) | done |
-| HigherHRNet (CVPR 2020) | |
-| DeepPose (CVPR 2014) | done |
-| RLE (ICCV 2021) | done |
-| SoftWingloss (TIP 2021) | |
-| VideoPose3D (CVPR 2019) | |
-| Hourglass (ECCV 2016) | done |
-| LiteHRNet (CVPR 2021) | done |
-| AdaptiveWingloss (ICCV 2019) | done |
-| SimpleBaseline2D (ECCV 2018) | done |
-| PoseWarper (NeurIPS 2019) | |
-| SimpleBaseline3D (ICCV 2017) | |
-| HMR (CVPR 2018) | |
-| UDP (CVPR 2020) | done |
-| VIPNAS (CVPR 2021) | done |
-| Wingloss (CVPR 2018) | |
-| DarkPose (CVPR 2020) | done |
+| Algorithm                         |   Status    |
+| :-------------------------------- | :---------: |
+| MTUT (CVPR 2019)                  |             |
+| MSPN (ArXiv 2019)                 |    done     |
+| InterNet (ECCV 2020)              |             |
+| DEKR (CVPR 2021)                  |    done     |
+| HigherHRNet (CVPR 2020)           |             |
+| DeepPose (CVPR 2014)              |    done     |
+| RLE (ICCV 2021)                   |    done     |
+| SoftWingloss (TIP 2021)           |             |
+| VideoPose3D (CVPR 2019)           |             |
+| Hourglass (ECCV 2016)             |    done     |
+| LiteHRNet (CVPR 2021)             |    done     |
+| AdaptiveWingloss (ICCV 2019)      |    done     |
+| SimpleBaseline2D (ECCV 2018)      |    done     |
+| PoseWarper (NeurIPS 2019)         |             |
+| SimpleBaseline3D (ICCV 2017)      |             |
+| HMR (CVPR 2018)                   |             |
+| UDP (CVPR 2020)                   |    done     |
+| VIPNAS (CVPR 2021)                |    done     |
+| Wingloss (CVPR 2018)              |             |
+| DarkPose (CVPR 2020)              |    done     |
 | Associative Embedding (NIPS 2017) | in progress |
-| VoxelPose (ECCV 2020) | |
-| RSN (ECCV 2020) | done |
-| CID (CVPR 2022) | done |
-| CPM (CVPR 2016) | done |
-| HRNet (CVPR 2019) | done |
-| HRNetv2 (TPAMI 2019) | done |
-| SCNet (CVPR 2020) | done |
+| VoxelPose (ECCV 2020)             |             |
+| RSN (ECCV 2020)                   |    done     |
+| CID (CVPR 2022)                   |    done     |
+| CPM (CVPR 2016)                   |    done     |
+| HRNet (CVPR 2019)                 |    done     |
+| HRNetv2 (TPAMI 2019)              |    done     |
+| SCNet (CVPR 2020)                 |    done     |
 
 </details>
 

README_CN.md

@@ -128,40 +128,41 @@ MMPose v1.0.0 是一个重大更新，包括了大量的 API 和配置文件的
 <details close>
 <summary><b>迁移进度</b></summary>
 
-| 算法名称 | 迁移进度 |
-| :--- | :---: |
-| MTUT (CVPR 2019) |  | 
-| MSPN (ArXiv 2019)| done | 
-| InterNet (ECCV 2020) |  | 
-| DEKR (CVPR 2021) | done |
-| HigherHRNet (CVPR 2020) | |
-| DeepPose (CVPR 2014) | done |
-| RLE (ICCV 2021) | done |
-| SoftWingloss (TIP 2021) | |
-| VideoPose3D (CVPR 2019) | |
-| Hourglass (ECCV 2016) | done |
-| LiteHRNet (CVPR 2021) | done |
-| AdaptiveWingloss (ICCV 2019) | done |
-| SimpleBaseline2D (ECCV 2018) | done |
-| PoseWarper (NeurIPS 2019) | |
-| SimpleBaseline3D (ICCV 2017) | |
-| HMR (CVPR 2018) | |
-| UDP (CVPR 2020) | done |
-| VIPNAS (CVPR 2021) | done |
-| Wingloss (CVPR 2018) | |
-| DarkPose (CVPR 2020) | done |
+| 算法名称                          |  迁移进度   |
+| :-------------------------------- | :---------: |
+| MTUT (CVPR 2019)                  |             |
+| MSPN (ArXiv 2019)                 |    done     |
+| InterNet (ECCV 2020)              |             |
+| DEKR (CVPR 2021)                  |    done     |
+| HigherHRNet (CVPR 2020)           |             |
+| DeepPose (CVPR 2014)              |    done     |
+| RLE (ICCV 2021)                   |    done     |
+| SoftWingloss (TIP 2021)           |             |
+| VideoPose3D (CVPR 2019)           |             |
+| Hourglass (ECCV 2016)             |    done     |
+| LiteHRNet (CVPR 2021)             |    done     |
+| AdaptiveWingloss (ICCV 2019)      |    done     |
+| SimpleBaseline2D (ECCV 2018)      |    done     |
+| PoseWarper (NeurIPS 2019)         |             |
+| SimpleBaseline3D (ICCV 2017)      |             |
+| HMR (CVPR 2018)                   |             |
+| UDP (CVPR 2020)                   |    done     |
+| VIPNAS (CVPR 2021)                |    done     |
+| Wingloss (CVPR 2018)              |             |
+| DarkPose (CVPR 2020)              |    done     |
 | Associative Embedding (NIPS 2017) | in progress |
-| VoxelPose (ECCV 2020) | |
-| RSN (ECCV 2020) | done |
-| CID (CVPR 2022) | done |
-| CPM (CVPR 2016) | done |
-| HRNet (CVPR 2019) | done |
-| HRNetv2 (TPAMI 2019) | done |
-| SCNet (CVPR 2020) | done |
+| VoxelPose (ECCV 2020)             |             |
+| RSN (ECCV 2020)                   |    done     |
+| CID (CVPR 2022)                   |    done     |
+| CPM (CVPR 2016)                   |    done     |
+| HRNet (CVPR 2019)                 |    done     |
+| HRNetv2 (TPAMI 2019)              |    done     |
+| SCNet (CVPR 2020)                 |    done     |
 
 </details>
 
 如果您使用的算法还没有完成迁移，您也可以继续使用访问 [0.x 分支](https://github.com/open-mmlab/mmpose/tree/0.x) 和 [旧版文档](https://mmpose.readthedocs.io/zh_CN/0.x/)
+
 ## 安装
 
 关于安装的详细说明请参考[安装文档](https://mmpose.readthedocs.io/zh_CN/latest/installation.html)。

configs/body_2d_keypoint/topdown_heatmap/coco/vitpose_coco.md

@@ -57,6 +57,6 @@ Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 da
 | Arch                                          | Input Size |  AP   | AP<sup>50</sup> | AP<sup>75</sup> |  AR   | AR<sup>50</sup> |                     ckpt                      |                      log                      |
 | :-------------------------------------------- | :--------: | :---: | :-------------: | :-------------: | :---: | :-------------: | :-------------------------------------------: | :-------------------------------------------: |
 | [ViTPose-S](/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-small-simple_8xb64-210e_coco-256x192.py) |  256x192   | 0.736 |      0.900      |      0.811      | 0.790 |      0.940      | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-small-simple_8xb64-210e_coco-256x192-4c101a76_20230314.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-small-simple_8xb64-210e_coco-256x192-4c101a76_20230314.json) |
-| [ViTPose-B](/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192.py) |  256x192   | 0.756 |      0.906      |      0.826      | 0.809 |      0.946      | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192-fd73707d_20230314.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192-fd73707d_20230314.json) |
+| [ViTPose-B](/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192.py) |  256x192   | 0.756 |      0.906      |      0.826      | 0.809 |      0.946      | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192-0b8234ea_20230407.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192-0b8234ea_20230407.json) |
 | [ViTPose-L](/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-large-simple_8xb64-210e_coco-256x192.py) |  256x192   | 0.781 |      0.914      |      0.853      | 0.833 |      0.952      | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-large-simple_8xb64-210e_coco-256x192-3a7ee9e1_20230314.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-large-simple_8xb64-210e_coco-256x192-3a7ee9e1_20230314.json) |
 | [ViTPose-H](/configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-huge-simple_8xb64-210e_coco-256x192.py) |  256x192   | 0.789 |      0.916      |      0.856      | 0.839 |      0.953      | [ckpt](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-huge-simple_8xb64-210e_coco-256x192-ffd48c05_20230314.pth) | [log](https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-huge-simple_8xb64-210e_coco-256x192-ffd48c05_20230314.json) |

configs/body_2d_keypoint/topdown_heatmap/coco/vitpose_coco.yml

@@ -116,7 +116,7 @@ Models:
       AR: 0.809
       AR@0.5: 0.946
     Task: Body 2D Keypoint
-  Weights: https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192-fd73707d_20230314.pth
+  Weights: https://download.openmmlab.com/mmpose/v1/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-base-simple_8xb64-210e_coco-256x192-0b8234ea_20230407.pth
 - Config: configs/body_2d_keypoint/topdown_heatmap/coco/td-hm_ViTPose-large-simple_8xb64-210e_coco-256x192.py
   In Collection: ViTPose
   Alias: vitpose-l

mmpose/datasets/datasets/base/base_coco_style_dataset.py

@@ -7,8 +7,8 @@
 
 import numpy as np
 from mmengine.dataset import BaseDataset, force_full_init
-from mmengine.fileio import load
-from mmengine.utils import check_file_exist, is_list_of
+from mmengine.fileio import exists, get_local_path, load
+from mmengine.utils import is_list_of
 from xtcocotools.coco import COCO
 
 from mmpose.registry import DATASETS
@@ -195,18 +195,19 @@ def load_data_list(self) -> List[dict]:
     def _load_annotations(self) -> Tuple[List[dict], List[dict]]:
         """Load data from annotations in COCO format."""
 
-        check_file_exist(self.ann_file)
+        assert exists(self.ann_file), 'Annotation file does not exist'
 
-        coco = COCO(self.ann_file)
+        with get_local_path(self.ann_file) as local_path:
+            self.coco = COCO(local_path)
         # set the metainfo about categories, which is a list of dict
         # and each dict contains the 'id', 'name', etc. about this category
-        self._metainfo['CLASSES'] = coco.loadCats(coco.getCatIds())
+        self._metainfo['CLASSES'] = self.coco.loadCats(self.coco.getCatIds())
 
         instance_list = []
         image_list = []
 
-        for img_id in coco.getImgIds():
-            img = coco.loadImgs(img_id)[0]
+        for img_id in self.coco.getImgIds():
+            img = self.coco.loadImgs(img_id)[0]
             img.update({
                 'img_id':
                 img_id,
@@ -215,8 +216,8 @@ def _load_annotations(self) -> Tuple[List[dict], List[dict]]:
             })
             image_list.append(img)
 
-            ann_ids = coco.getAnnIds(imgIds=img_id)
-            for ann in coco.loadAnns(ann_ids):
+            ann_ids = self.coco.getAnnIds(imgIds=img_id)
+            for ann in self.coco.loadAnns(ann_ids):
 
                 instance_info = self.parse_data_info(
                     dict(raw_ann_info=ann, raw_img_info=img))
@@ -380,18 +381,18 @@ def _get_bottomup_data_infos(self, instance_list: List[Dict],
     def _load_detection_results(self) -> List[dict]:
         """Load data from detection results with dummy keypoint annotations."""
 
-        check_file_exist(self.ann_file)
-        check_file_exist(self.bbox_file)
-
+        assert exists(self.ann_file), 'Annotation file does not exist'
+        assert exists(self.bbox_file), 'Bbox file does not exist'
         # load detection results
         det_results = load(self.bbox_file)
         assert is_list_of(det_results, dict)
 
         # load coco annotations to build image id-to-name index
-        coco = COCO(self.ann_file)
+        with get_local_path(self.ann_file) as local_path:
+            self.coco = COCO(local_path)
         # set the metainfo about categories, which is a list of dict
         # and each dict contains the 'id', 'name', etc. about this category
-        self._metainfo['CLASSES'] = coco.loadCats(coco.getCatIds())
+        self._metainfo['CLASSES'] = self.coco.loadCats(self.coco.getCatIds())
 
         num_keypoints = self.metainfo['num_keypoints']
         data_list = []
@@ -401,7 +402,7 @@ def _load_detection_results(self) -> List[dict]:
             if det['category_id'] != 1:
                 continue
 
-            img = coco.loadImgs(det['image_id'])[0]
+            img = self.coco.loadImgs(det['image_id'])[0]
 
             img_path = osp.join(self.data_prefix['img'], img['file_name'])
             bbox_xywh = np.array(

mmpose/datasets/datasets/body/mpii_dataset.py

@@ -4,7 +4,7 @@
 from typing import Callable, List, Optional, Sequence, Tuple, Union
 
 import numpy as np
-from mmengine.utils import check_file_exist
+from mmengine.fileio import exists, get_local_path
 from scipy.io import loadmat
 
 from mmpose.registry import DATASETS
@@ -137,14 +137,16 @@ def __init__(self,
     def _load_annotations(self) -> Tuple[List[dict], List[dict]]:
         """Load data from annotations in MPII format."""
 
-        check_file_exist(self.ann_file)
-        with open(self.ann_file) as anno_file:
-            anns = json.load(anno_file)
+        assert exists(self.ann_file), 'Annotation file does not exist'
+        with get_local_path(self.ann_file) as local_path:
+            with open(local_path) as anno_file:
+                self.anns = json.load(anno_file)
 
         if self.headbox_file:
-            check_file_exist(self.headbox_file)
-            headbox_dict = loadmat(self.headbox_file)
-            headboxes_src = np.transpose(headbox_dict['headboxes_src'],
+            assert exists(self.headbox_file), 'Headbox file does not exist'
+            with get_local_path(self.headbox_file) as local_path:
+                self.headbox_dict = loadmat(local_path)
+            headboxes_src = np.transpose(self.headbox_dict['headboxes_src'],
                                          [2, 0, 1])
             SC_BIAS = 0.6
 
@@ -156,7 +158,7 @@ def _load_annotations(self) -> Tuple[List[dict], List[dict]]:
         # mpii bbox scales are normalized with factor 200.
         pixel_std = 200.
 
-        for idx, ann in enumerate(anns):
+        for idx, ann in enumerate(self.anns):
             center = np.array(ann['center'], dtype=np.float32)
             scale = np.array([ann['scale'], ann['scale']],
                              dtype=np.float32) * pixel_std

mmpose/datasets/datasets/body/mpii_trb_dataset.py

@@ -4,7 +4,7 @@
 from typing import List, Tuple
 
 import numpy as np
-from mmengine.utils import check_file_exist
+from mmengine.fileio import exists, get_local_path
 
 from mmpose.registry import DATASETS
 from mmpose.structures.bbox import bbox_cs2xyxy
@@ -106,11 +106,12 @@ class MpiiTrbDataset(BaseCocoStyleDataset):
     def _load_annotations(self) -> Tuple[List[dict], List[dict]]:
         """Load data from annotations in MPII-TRB format."""
 
-        check_file_exist(self.ann_file)
-        with open(self.ann_file) as anno_file:
-            data = json.load(anno_file)
+        assert exists(self.ann_file), 'Annotation file does not exist'
+        with get_local_path(self.ann_file) as local_path:
+            with open(local_path) as anno_file:
+                self.data = json.load(anno_file)
 
-        imgid2info = {img['id']: img for img in data['images']}
+        imgid2info = {img['id']: img for img in self.data['images']}
 
         instance_list = []
         image_list = []
@@ -119,7 +120,7 @@ def _load_annotations(self) -> Tuple[List[dict], List[dict]]:
         # mpii-trb bbox scales are normalized with factor 200.
         pixel_std = 200.
 
-        for ann in data['annotations']:
+        for ann in self.data['annotations']:
             img_id = ann['image_id']
 
             # center, scale in shape [1, 2] and bbox in [1, 4]

mmpose/datasets/datasets/body/posetrack18_video_dataset.py

@@ -3,8 +3,8 @@
 from typing import Callable, List, Optional, Sequence, Union
 
 import numpy as np
-from mmengine.fileio import load
-from mmengine.utils import check_file_exist, is_list_of
+from mmengine.fileio import exists, get_local_path, load
+from mmengine.utils import is_list_of
 from xtcocotools.coco import COCO
 
 from mmpose.registry import DATASETS
@@ -287,22 +287,22 @@ def parse_data_info(self, raw_data_info: dict) -> Optional[dict]:
 
     def _load_detection_results(self) -> List[dict]:
         """Load data from detection results with dummy keypoint annotations."""
-
-        check_file_exist(self.ann_file)
-        check_file_exist(self.bbox_file)
+        assert exists(self.ann_file), 'Annotation file does not exist'
+        assert exists(self.bbox_file), 'Bbox file does not exist'
 
         # load detection results
         det_results = load(self.bbox_file)
         assert is_list_of(det_results, dict)
 
         # load coco annotations to build image id-to-name index
-        coco = COCO(self.ann_file)
+        with get_local_path(self.ann_file) as local_path:
+            self.coco = COCO(local_path)
 
         # mapping image name to id
         name2id = {}
         # mapping image id to name
         id2name = {}
-        for img_id, image in coco.imgs.items():
+        for img_id, image in self.coco.imgs.items():
             file_name = image['file_name']
             id2name[img_id] = file_name
             name2id[file_name] = img_id
@@ -333,7 +333,7 @@ def _load_detection_results(self) -> List[dict]:
                     img_id = name2id[det['image_name']]
                 else:
                     img_id = det['image_id']
-                img_ann = coco.loadImgs(img_id)[0]
+                img_ann = self.coco.loadImgs(img_id)[0]
                 nframes = int(img_ann['nframes'])
 
             # deal with multiple image paths

mmpose/datasets/datasets/hand/coco_wholebody_hand_dataset.py

@@ -3,7 +3,7 @@
 from typing import List, Tuple
 
 import numpy as np
-from mmengine.utils import check_file_exist
+from mmengine.fileio import exists, get_local_path
 from xtcocotools.coco import COCO
 
 from mmpose.registry import DATASETS
@@ -87,15 +87,16 @@ class CocoWholeBodyHandDataset(BaseCocoStyleDataset):
     def _load_annotations(self) -> Tuple[List[dict], List[dict]]:
         """Load data from annotations in COCO format."""
 
-        check_file_exist(self.ann_file)
+        assert exists(self.ann_file), 'Annotation file does not exist'
 
-        coco = COCO(self.ann_file)
+        with get_local_path(self.ann_file) as local_path:
+            self.coco = COCO(local_path)
         instance_list = []
         image_list = []
         id = 0
 
-        for img_id in coco.getImgIds():
-            img = coco.loadImgs(img_id)[0]
+        for img_id in self.coco.getImgIds():
+            img = self.coco.loadImgs(img_id)[0]
 
             img.update({
                 'img_id':
@@ -105,8 +106,8 @@ def _load_annotations(self) -> Tuple[List[dict], List[dict]]:
             })
             image_list.append(img)
 
-            ann_ids = coco.getAnnIds(imgIds=img_id, iscrowd=False)
-            anns = coco.loadAnns(ann_ids)
+            ann_ids = self.coco.getAnnIds(imgIds=img_id, iscrowd=False)
+            anns = self.coco.loadAnns(ann_ids)
             for ann in anns:
                 for type in ['left', 'right']:
                     # filter invalid hand annotations, there might be two