fix linting

mmhuman3d/data/data_converters/aic.py

@@ -13,10 +13,10 @@
 
 @DATA_CONVERTERS.register_module()
 class AicConverter(BaseModeConverter):
-    """AI Challenger dataset
-    `Ai challenger: A large-scale dataset for going deeper in image 
-    understanding' arXiv'2017
-    More details can be found in the `paper
+    """AI Challenger dataset `Ai challenger: A large-scale dataset for going
+    deeper in image understanding' arXiv'2017 More details can be found in the
+    `paper.
+
     <https://arxiv.org/abs/1711.06475>`__ .
 
     Args:
@@ -88,8 +88,7 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
         bbox_xywh_ = np.array(bbox_xywh_).reshape((-1, 4))
         bbox_xywh_ = np.hstack([bbox_xywh_, np.ones([bbox_xywh_.shape[0], 1])])
         keypoints2d_ = np.array(keypoints2d_).reshape((-1, 14, 3))
-        keypoints2d_, mask = convert_kps(keypoints2d_, 'aic',
-                                         'human_data')
+        keypoints2d_, mask = convert_kps(keypoints2d_, 'aic', 'human_data')
 
         human_data['image_path'] = image_path_
         human_data['keypoints2d_mask'] = mask

mmhuman3d/data/data_converters/instavariety.py

@@ -1,13 +1,11 @@
-from .base_converter import BaseModeConverter
-from typing import List
-import json
-import os
 import glob
+import os
+from typing import List
 
 import cv2
 import numpy as np
-from tqdm import tqdm
 import tensorflow as tf
+from tqdm import tqdm
 
 from mmhuman3d.core.conventions.keypoints_mapping import convert_kps
 from mmhuman3d.data.data_structures.human_data import HumanData
@@ -28,9 +26,7 @@ class InstaVarietyConverter(BaseModeConverter):
     """
     ACCEPTED_MODES = ['train', 'test']
 
-    def __init__(self,
-                 modes: List = [],
-                 extract_img: bool = False) -> None:
+    def __init__(self, modes: List = [], extract_img: bool = False) -> None:
         super(InstaVarietyConverter, self).__init__(modes)
         self.extract_img = extract_img
 
@@ -86,13 +82,13 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
 
             toe_pts = np.array(toe_pts).reshape(-1, 3, 6)
 
-            visibles = example.features.feature[
+            vis = example.features.feature[
                 'image/visibilities'].int64_list.value
-            visibles = np.array(visibles).reshape(-1, 1, 14)
+            vis = np.array(vis).reshape(-1, 1, 14)
 
             for i in tqdm(range(N)):
                 image = tf.image.decode_jpeg(images_data[i], channels=3)
-                kp = np.vstack((xys[i], visibles[i]))
+                kp = np.vstack((xys[i], vis[i]))
                 faces = face_pts[i]
 
                 toes = toe_pts[i]
@@ -118,22 +114,20 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
                 bbox_xyxy = self._bbox_expand(bbox_xyxy, scale_factor=1.2)
                 bbox_xywh = self._xyxy2xywh(bbox_xyxy)
 
-                image_path = images_name[i].decode(
-                    "utf-8").replace('/data2/Data/instagram_download/frames_raw/', 'images/')
-
+                image_path = images_name[i].decode('utf-8').replace(
+                    '/data2/Data/instagram_download/frames_raw/', 'images/')
 
                 if self.extract_img:
                     image_abs_path = os.path.join(dataset_path, image_path)
                     folder = os.path.dirname(image_abs_path)
                     if not os.path.exists(folder):
                         os.makedirs(folder, exist_ok=True)
                     cv2.imwrite(image_abs_path, np.array(image))
-                    
+
                 image_path_.append(image_path)
                 keypoints2d_.append(keypoints2d)
                 bbox_xywh_.append(bbox_xywh)
 
-
         # convert keypoints
         bbox_xywh_ = np.array(bbox_xywh_).reshape((-1, 4))
         bbox_xywh_ = np.hstack([bbox_xywh_, np.ones([bbox_xywh_.shape[0], 1])])
@@ -152,4 +146,3 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
             os.makedirs(out_path)
         out_file = os.path.join(out_path, f'instavariety_{mode}.npz')
         human_data.dump(out_file)
-

mmhuman3d/data/data_converters/mtp.py

@@ -1,12 +1,10 @@
+import json
 import os
 import pickle
-import xml.etree.ElementTree as ET
 from typing import List
 
 import numpy as np
 from tqdm import tqdm
-import json
-import pickle
 
 from mmhuman3d.core.conventions.keypoints_mapping import convert_kps
 from mmhuman3d.data.data_structures.human_data import HumanData
@@ -16,18 +14,17 @@
 
 @DATA_CONVERTERS.register_module()
 class MtpConverter(BaseModeConverter):
-    """MTP dataset
-    `On Self-Contact and Human Pose' CVPR`2021
-    More details can be found in the `paper
+    """MTP dataset `On Self-Contact and Human Pose' CVPR`2021 More details can
+    be found in the `paper.
+
     <https://arxiv.org/pdf/2104.03176.pdf>`__.
 
     Args:
         modes (list): 'valid' or 'train' for accepted modes
     """
     ACCEPTED_MODES = ['train', 'val']
 
-    def __init__(self,
-                 modes: List = []) -> None:
+    def __init__(self, modes: List = []) -> None:
         super(MtpConverter, self).__init__(modes)
 
     def convert_by_mode(self, dataset_path: str, out_path: str,
@@ -63,17 +60,23 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
         for img_id in tqdm(json_data):
             part_name = img_id.split('_')[1]
             image_path = f'images/{part_name}/{img_id}.png'
-            keypoints_file = os.path.join(dataset_path, f'keypoints/openpose/{part_name}/{img_id}.json')
-            smpl_file = os.path.join(dataset_path, f'smplify-xmc/smpl/params/{part_name}/{img_id}.pkl')
+            keypoints_file = os.path.join(
+                dataset_path, f'keypoints/openpose/{part_name}/{img_id}.json')
+            smpl_file = os.path.join(
+                dataset_path,
+                f'smplify-xmc/smpl/params/{part_name}/{img_id}.pkl')
 
             with open(keypoints_file) as f:
                 data = json.load(f)
-                if len(data['people']) <1 :
+                if len(data['people']) < 1:
                     continue
-                keypoints2d = np.array(data['people'][0]['pose_keypoints_2d']).reshape(25, 3)
-                keypoints2d[keypoints2d[:, 2] > 0.15, 2] = 1 # set based on keypoints confidence 
-
-                vis_keypoints2d = keypoints2d[np.where(keypoints2d[:, 2]>0)[0]] 
+                keypoints2d = np.array(
+                    data['people'][0]['pose_keypoints_2d']).reshape(25, 3)
+                keypoints2d[keypoints2d[:, 2] > 0.15,
+                            2] = 1  # set based on keypoints confidence
+
+                vis_keypoints2d = keypoints2d[np.where(
+                    keypoints2d[:, 2] > 0)[0]]
                 # bbox
                 bbox_xyxy = [
                     min(vis_keypoints2d[:, 0]),
@@ -95,12 +98,11 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
             keypoints2d_.append(keypoints2d)
             bbox_xywh_.append(bbox_xywh)
 
-        smpl['body_pose'] = np.array(smpl['body_pose']).reshape(
-            (-1, 23, 3))
+        smpl['body_pose'] = np.array(smpl['body_pose']).reshape((-1, 23, 3))
         smpl['global_orient'] = np.array(smpl['global_orient']).reshape(
             (-1, 3))
         smpl['betas'] = np.array(smpl['betas']).reshape((-1, 10))
-        
+
         bbox_xywh_ = np.array(bbox_xywh_).reshape((-1, 4))
         bbox_xywh_ = np.hstack([bbox_xywh_, np.ones([bbox_xywh_.shape[0], 1])])
         keypoints2d_ = np.array(keypoints2d_).reshape((-1, 25, 3))
@@ -121,4 +123,3 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
         file_name = 'mtp_{}.npz'.format(mode)
         out_file = os.path.join(out_path, file_name)
         human_data.dump(out_file)
-

mmhuman3d/data/data_converters/muco3dhp.py

@@ -15,12 +15,12 @@
 
 @DATA_CONVERTERS.register_module()
 class Muco3dhpConverter(BaseConverter):
-    """MuCo-3DHP dataset `Single-Shot Multi-Person 3D Pose Estimation 
-    From Monocular RGB' 3DV'2018
-    More details can be found in the `paper.
+    """MuCo-3DHP dataset `Single-Shot Multi-Person 3D Pose Estimation From
+    Monocular RGB' 3DV'2018 More details can be found in the `paper.
 
     <https://arxiv.org/abs/1712.03453>`__ .
     """
+
     @staticmethod
     def get_intrinsic_matrix(f: List[float],
                              c: List[float],
@@ -90,8 +90,7 @@ def convert(self, dataset_path: str, out_path: str) -> dict:
                     smpl_param = smpl_params[str(pid)]
                     pose, shape, trans = np.array(
                         smpl_param['pose']), np.array(
-                            smpl_param['shape']), np.array(
-                                smpl_param['trans'])
+                            smpl_param['shape']), np.array(smpl_param['trans'])
                     sum = pose.sum() + shape.sum() + trans.sum()
                     if np.isnan(sum):
                         continue
@@ -142,4 +141,4 @@ def convert(self, dataset_path: str, out_path: str) -> dict:
 
         file_name = 'muco3dhp_train.npz'
         out_file = os.path.join(out_path, file_name)
-        human_data.dump(out_file)
+        human_data.dump(out_file)

mmhuman3d/data/data_converters/mupots3d.py

@@ -1,9 +1,6 @@
-import glob
 import os
 from typing import List, Tuple
 
-import cv2
-import h5py
 import numpy as np
 import scipy.io as sio
 from tqdm import tqdm
@@ -16,32 +13,34 @@
 
 @DATA_CONVERTERS.register_module()
 class Mupots3dConverter(BaseConverter):
-    """MuPoTs-3D dataset `Single-Shot Multi-Person 3D Pose Estimation 
-    From Monocular RGB' 3DV'2018
-    More details can be found in the `paper.
+    """MuPoTs-3D dataset `Single-Shot Multi-Person 3D Pose Estimation From
+    Monocular RGB' 3DV'2018 More details can be found in the `paper.
 
     <https://arxiv.org/abs/1712.03453>`__ .
     """
+
     @staticmethod
     def load_annot(fname):
+
         def parse_pose(dt):
             res = {}
-            annot2 = dt['annot2'][0,0]
-            annot3 = dt['annot3'][0,0]
-            annot3_univ = dt['univ_annot3'][0,0]
-            is_valid = dt['isValidFrame'][0,0][0,0]
+            annot2 = dt['annot2'][0, 0]
+            annot3 = dt['annot3'][0, 0]
+            annot3_univ = dt['univ_annot3'][0, 0]
+            is_valid = dt['isValidFrame'][0, 0][0, 0]
             res['annot2'] = annot2
             res['annot3'] = annot3
             res['annot3_univ'] = annot3_univ
             res['is_valid'] = is_valid
-            return res 
+            return res
+
         data = sio.loadmat(fname)['annotations']
         results = []
         num_frames, num_inst = data.shape[0], data.shape[1]
         for j in range(num_inst):
             buff = []
             for i in range(num_frames):
-                buff.append(parse_pose(data[i,j]))
+                buff.append(parse_pose(data[i, j]))
             results.append(buff)
         return results
 
@@ -94,7 +93,6 @@ def convert(self, dataset_path: str, out_path: str) -> dict:
 
         image_path_, bbox_xywh_, keypoints2d_, keypoints3d_ = [], [], [], []
 
-
         # test data
         user_list = range(1, 21)
 
@@ -131,14 +129,13 @@ def convert(self, dataset_path: str, out_path: str) -> dict:
                     keypoints3d_.append(keypoints3d)
 
         bbox_xywh_ = np.array(bbox_xywh_).reshape((-1, 4))
-        bbox_xywh_ = np.hstack(
-            [bbox_xywh_, np.ones([bbox_xywh_.shape[0], 1])])
+        bbox_xywh_ = np.hstack([bbox_xywh_, np.ones([bbox_xywh_.shape[0], 1])])
         keypoints2d_ = np.array(keypoints2d_).reshape((-1, 17, 3))
         keypoints2d_, mask = convert_kps(keypoints2d_, 'mpi_inf_3dhp_test',
-                                            'human_data')
+                                         'human_data')
         keypoints3d_ = np.array(keypoints3d_).reshape((-1, 17, 4))
         keypoints3d_, _ = convert_kps(keypoints3d_, 'mpi_inf_3dhp_test',
-                                        'human_data')
+                                      'human_data')
 
         human_data['image_path'] = image_path_
         human_data['bbox_xywh'] = bbox_xywh_

mmhuman3d/data/data_converters/ochuman.py

@@ -81,8 +81,7 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
         bbox_xywh_ = np.array(bbox_xywh_).reshape((-1, 4))
         bbox_xywh_ = np.hstack([bbox_xywh_, np.ones([bbox_xywh_.shape[0], 1])])
         keypoints2d_ = np.array(keypoints2d_).reshape((-1, 19, 3))
-        keypoints2d_, mask = convert_kps(keypoints2d_, 'ochuman',
-                                         'human_data')
+        keypoints2d_, mask = convert_kps(keypoints2d_, 'ochuman', 'human_data')
 
         human_data['image_path'] = image_path_
         human_data['keypoints2d_mask'] = mask

mmhuman3d/data/data_converters/oh50k3d.py

@@ -1,8 +1,8 @@
 import json
 import os
-import cv2
 from typing import List
 
+import cv2
 import numpy as np
 from tqdm import tqdm
 
@@ -15,10 +15,9 @@
 
 @DATA_CONVERTERS.register_module()
 class OH50k3DConverter(BaseModeConverter):
-    """3DOH50K dataset
-    `Object-Occluded Human Shape and Pose Estimation from a Single Color 
-    Image' CVPR'2020
-    More details can be found in the `paper
+    """3DOH50K dataset `Object-Occluded Human Shape and Pose Estimation from a
+    Single Color Image' CVPR'2020 More details can be found in the `paper.
+
     <https://www.yangangwang.com/papers/ZHANG-OOH-2020-03.pdf>`__ .
 
     Args:
@@ -48,7 +47,8 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
         human_data = HumanData()
 
         # structs we use
-        image_path_, bbox_xywh_, keypoints2d_, keypoints3d_, cam_param_ = [], [], [], [], []
+        image_path_, bbox_xywh_, keypoints2d_, keypoints3d_, cam_param_ \
+            = [], [], [], [], []
 
         smpl = {}
         smpl['body_pose'] = []
@@ -70,11 +70,10 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
             betas = np.array(annot['betas']).reshape(-1)
             pose = np.array(annot['pose']).reshape(-1)
             trans = np.array(annot['trans']).reshape(-1)
-            scale = np.array(annot['scale'])
-            smpl_joints_2d = np.array(annot['smpl_joints_2d']) # 24x2
-            smpl_joints_3d = np.array(annot['smpl_joints_3d']) # 24x3
-            lsp_joints_2d = np.array(annot['lsp_joints_2d']) # 14x2
-            lsp_joints_3d = np.array(annot['lsp_joints_3d']) # 14x3
+            smpl_joints_2d = np.array(annot['smpl_joints_2d'])  # 24x2
+            smpl_joints_3d = np.array(annot['smpl_joints_3d'])  # 24x3
+            lsp_joints_2d = np.array(annot['lsp_joints_2d'])  # 14x2
+            lsp_joints_3d = np.array(annot['lsp_joints_3d'])  # 14x3
 
             # fix keypoints3d
             smpl_joints_3d = smpl_joints_3d - smpl_joints_3d[0]
@@ -84,7 +83,7 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
             h, w, _ = cv2.imread(f'{dataset_path}/{img_path}').shape
 
             # scale and center
-            bbox_xyxy = np.array(annot['bbox']).reshape(-1) # 2x2 - check foramt
+            bbox_xyxy = np.array(annot['bbox']).reshape(-1)
             bbox_xyxy = self._bbox_expand(bbox_xyxy, scale_factor=1.2)
             bbox_xywh = self._xyxy2xywh(bbox_xyxy)
             smpl_joints_2d = np.hstack([smpl_joints_2d, np.ones([24, 1])])
@@ -98,9 +97,9 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
             camera = CameraParameter(H=h, W=w)
             camera.set_KRT(K, R, T)
             parameter_dict = camera.to_dict()
-            pose[:3] = cv2.Rodrigues(
-                np.dot(R,
-                        cv2.Rodrigues(pose[:3])[0]))[0].T[0]
+            pose[:3] = cv2.Rodrigues(np.dot(R,
+                                            cv2.Rodrigues(
+                                                pose[:3])[0]))[0].T[0]
 
             # store data
             image_path_.append(img_path)
@@ -119,13 +118,11 @@ def convert_by_mode(self, dataset_path: str, out_path: str,
         smpl['betas'] = np.array(smpl['betas']).reshape((-1, 10))
         smpl['transl'] = np.array(smpl['transl']).reshape((-1, 3))
 
-
         # convert keypoints
         bbox_xywh_ = np.array(bbox_xywh_).reshape((-1, 4))
         bbox_xywh_ = np.hstack([bbox_xywh_, np.ones([bbox_xywh_.shape[0], 1])])
         keypoints2d_ = np.array(keypoints2d_).reshape((-1, 24, 3))
-        keypoints2d_, mask = convert_kps(keypoints2d_, 'smpl',
-                                         'human_data')
+        keypoints2d_, mask = convert_kps(keypoints2d_, 'smpl', 'human_data')
         keypoints3d_ = np.array(keypoints3d_).reshape((-1, 24, 4))
         keypoints3d_, _ = convert_kps(keypoints3d_, 'smpl', 'human_data')