atten

Author: argusswift

config/yolov4_config.py

@@ -4,8 +4,6 @@
 PROJECT_PATH = osp.abspath(osp.join(osp.dirname(__file__), '..'))
 
 DATA_PATH = osp.join(PROJECT_PATH, 'data')
-# PROJECT_PATH = "E:\YOLOV4/data"
-# PROJECT_PATH = "E:\YOLOV4/"
 
 
 MODEL_TYPE = {

eval/evaluator.py

@@ -122,7 +122,7 @@ def __predict(self, img, test_shape, valid_scale, mode):
         with torch.no_grad():
             start_time = current_milli_time()
             if self.showatt:
-                _, p_d, beta = self.model(img)
+                _, p_d, atten = self.model(img)
             else:
                 _, p_d = self.model(img)
             self.inference_time += current_milli_time() - start_time
@@ -131,7 +131,7 @@ def __predict(self, img, test_shape, valid_scale, mode):
             pred_bbox, test_shape, (org_h, org_w), valid_scale
         )
         if self.showatt and len(img) and mode == 'det':
-            self.__show_heatmap(beta, org_img)
+            self.__show_heatmap(atten, org_img)
         return bboxes
 
     def __show_heatmap(self, beta, img):

model/YOLOv4.py

@@ -278,14 +278,14 @@ def __init__(self, weight_path=None, out_channels=255, resume=False, showatt=Fal
         self.predict_net = PredictNet(feature_channels, out_channels)
 
     def forward(self, x):
-        beta = None
+        atten = None
         features = self.backbone(x)
         if self.showatt:
-            features[-1], beta = self.attention(features[-1])
+            features[-1], atten = self.attention(features[-1])
         features[-1] = self.spp(features[-1])
         features = self.panet(features)
         predicts = self.predict_net(features)
-        return predicts, beta
+        return predicts, atten
 
 
 if __name__ == "__main__":

model/build_model.py

@@ -48,7 +48,7 @@ def __init__(self, weight_path=None, resume=False, showatt=False):
 
     def forward(self, x):
         out = []
-        [x_s, x_m, x_l], beta = self.__yolov4(x)
+        [x_s, x_m, x_l], atten = self.__yolov4(x)
 
         out.append(self.__head_s(x_s))
         out.append(self.__head_m(x_m))
@@ -60,7 +60,7 @@ def forward(self, x):
         else:
             p, p_d = list(zip(*out))
             if self.__showatt:
-                return p, torch.cat(p_d, 0), beta
+                return p, torch.cat(p_d, 0), atten
             return p, torch.cat(p_d, 0)
 
 

model/layers/global_context_block.py

@@ -48,7 +48,7 @@ def spatial_pool(self, x):
         context_mask = self.softmax(context_mask)
         beta1 = context_mask
         beta2 = torch.transpose(beta1, 1, 2)
-        beta = torch.matmul(beta2, beta1)
+        atten = torch.matmul(beta2, beta1)
 
         # [N, 1, H * W, 1]
         context_mask = context_mask.unsqueeze(3)
@@ -57,13 +57,13 @@ def spatial_pool(self, x):
         # [N, C, 1, 1]
         context = context.view(batch, channel, 1, 1)
 
-        return context, beta
+        return context, atten
 
     def forward(self, x):
         # [N, C, 1, 1]
-        context, beta = self.spatial_pool(x)
+        context, atten = self.spatial_pool(x)
         # [N, C, 1, 1]
         channel_add_term = self.channel_add_conv(context)
         out = x + channel_add_term
 
-        return out, beta
+        return out, atten