进一步修改eval

main.py

@@ -150,10 +150,10 @@ def net_eval():
         batch_msk_lst.append(msk_)
         bi += 1
         if bi == 4:
-            batch_res = eval_batch_scales(args, eval_net, batch_img_lst, scales=args.scales,
-                                          base_crop_size=args.crop_size, flip=args.flip)
-            for mi in range(args.batch_size):
-                hist += cal_hist(batch_msk_lst[mi].flatten(), batch_res[mi].flatten(), args.num_classes)
+            batch_res = eval_batch_scales(eval_net, batch_img_lst, scales=1.0,
+                                          base_crop_size=513, flip=True)
+            for mi in range(16):
+                hist += cal_hist(batch_msk_lst[mi].flatten(), batch_res[mi].flatten(), 6)
 
             bi = 0
             batch_img_lst = []
@@ -161,10 +161,10 @@ def net_eval():
             print('processed {} images'.format(i + 1))
         image_num = i
     if bi > 0:
-        batch_res = eval_batch_scales(args, eval_net, batch_img_lst, scales=args.scales,
-                                      base_crop_size=args.crop_size, flip=args.flip)
+        batch_res = eval_batch_scales(eval_net, batch_img_lst, scales=1.0,
+                                      base_crop_size=513, flip=True)
         for mi in range(bi):
-            hist += cal_hist(batch_msk_lst[mi].flatten(), batch_res[mi].flatten(), args.num_classes)
+            hist += cal_hist(batch_msk_lst[mi].flatten(), batch_res[mi].flatten(), 6)
         print('processed {} images'.format(image_num + 1))