Merge pull request deepinsight#2070 from maldil/master

[Refactoring] making the code more Pythonic

body/human_pose/ambiguity_aware/lib/core/function1.py

@@ -319,9 +319,8 @@ def evaluate(val_loader, pose_model, config, is_train=False):
             factor = 0.680019 if prefix == "h36m" else 0.7577316
     if not use_same_norm_3d: 
         factor_filename = f"../data/{prefix}_test_factor_3d.pkl" if not is_train else f"../data/{prefix}_train_factor_3d.pkl"
-        f = open(factor_filename, "rb") 
-        factor = torch.from_numpy(pkl.load(f)).cuda()
-        f.close()
+        with open(factor_filename, "rb") as f:
+            factor = torch.from_numpy(pkl.load(f)).cuda()
     # print('eval with 15 joints')
     # exclude spine: 14, neck: 15
     # joint_indices = list(range(14)) + [16]

body/human_pose/ambiguity_aware/lib/core/function2.py

@@ -318,9 +318,8 @@ def evaluate(val_loader, pose_model, config, is_train=False):
         factor = 0.680019 if prefix == "h36m" else 0.7577316
     if not use_same_norm_3d: 
         factor_filename = f"../data/{prefix}_test_factor_3d.pkl" if not is_train else f"../data/{prefix}_train_factor_3d.pkl"
-        f = open(factor_filename, "rb") 
-        factor = torch.from_numpy(pkl.load(f)).cuda()
-        f.close()
+        with open(factor_filename, "rb") as f:
+            factor = torch.from_numpy(pkl.load(f)).cuda()
     # print('eval with 15 joints')
     # joint_indices = list(range(14)) + [16]
     with torch.no_grad():

body/human_pose/ambiguity_aware/lib/dataloader/mpiinf.py

@@ -75,9 +75,8 @@ def _load_data_set(self):
                 pkl.dump(factor_3d, f)
 
         if osp.exists(self.scale_path):
-            f = open(self.scale_path, "rb")
-            self.scales = pkl.load(f)['scale']
-            f.close()
+            with open(self.scale_path, "rb") as f:
+                self.scales = pkl.load(f)['scale']
         else:
             if self.use_scaler:
                 pass
@@ -86,9 +85,8 @@ def _load_data_set(self):
 
         if self.use_ideal_scale: 
             # scales computed from projection of 3d 
-            f = open("../data/mpi_{}_scales.pkl".format("train" if self.is_train else "valid"), "rb")
-            scales = pkl.load(f)
-            f.close()
+            with open("../data/mpi_{}_scales.pkl".format("train" if self.is_train else "valid"), "rb") as f:
+                scales = pkl.load(f)
             self.kp2ds = self.kp2ds * scales
 
         fp.close()

body/human_pose/ambiguity_aware/lib/utils/misc.py

@@ -48,15 +48,13 @@ def init_weights(model):
             m.bias.data.zero_()
 
 def save_pickle(data, save_path):
-    f = open(save_path, "wb")
-    pkl.dump(data, f)
-    f.close()
+    with open(save_path, "wb") as f:
+        pkl.dump(data, f)
     print(f"=> saved to {save_path}")
 
 def load_pickle(load_path):
-    f = open(load_path, "rb")
-    data = pkl.load(f)
-    f.close()
+    with open(load_path, "rb") as f:
+        data = pkl.load(f)
     print(f"<= loaded from {load_path}")
     return data 
 

detection/scrfd/mmdet/core/evaluation/widerface.py

@@ -153,9 +153,8 @@ def get_gt_boxes_from_txt(gt_path, cache_dir):
 
     cache_file = os.path.join(cache_dir, 'gt_cache.pkl')
     if os.path.exists(cache_file):
-        f = open(cache_file, 'rb')
-        boxes = pickle.load(f)
-        f.close()
+        with open(cache_file, 'rb') as f:
+            boxes = pickle.load(f)
         return boxes
 
     f = open(gt_path, 'r')
@@ -188,9 +187,8 @@ def get_gt_boxes_from_txt(gt_path, cache_dir):
             current_boxes.append(box)
             continue
 
-    f = open(cache_file, 'wb')
-    pickle.dump(boxes, f)
-    f.close()
+    with open(cache_file, 'wb') as f:
+        pickle.dump(boxes, f)
     return boxes
 
 

detection/scrfd/mmdet/models/necks/bfp.py

@@ -2,7 +2,7 @@
 import torch.nn.functional as F
 from mmcv.cnn import ConvModule, xavier_init
 from mmcv.cnn.bricks import NonLocal2d
-
+import numpy as np
 from ..builder import NECKS
 
 
@@ -85,7 +85,7 @@ def forward(self, inputs):
                     inputs[i], size=gather_size, mode='nearest')
             feats.append(gathered)
 
-        bsf = sum(feats) / len(feats)
+        bsf = np.mean(feats)
 
         # step 2: refine gathered features
         if self.refine_type is not None: