修改net_eval函数

main.py

@@ -42,7 +42,6 @@ def cal_hist(a, b, n):
     return np.bincount(n * a[k].astype(np.int32) + b[k], minlength=n ** 2).reshape(n, n)
 
 def resize_long(img, long_size=513):
-    print(img)
     h, w, _ = img.shape
     if h > w:
         new_h = long_size
@@ -93,9 +92,7 @@ def eval_batch(eval_net, img_lst, crop_size=513, flip=True):
     batch_img = np.zeros((4, 3, crop_size, crop_size), dtype=np.float32)
     resize_hw = []
     for l in range(batch_size):
-        print(img_lst)
         img_ = img_lst[l]
-        print(img_)
         img_, resize_h, resize_w = pre_process(img_, crop_size)
         batch_img[l] = img_
         resize_hw.append([resize_h, resize_w])
@@ -151,14 +148,13 @@ def net_eval():
         msk_path = './datasets/SegmentationClass/' + line.replace('\n', '') + '.png'
         img_ = cv2.imread(img_path)
         msk_ = cv2.imread(msk_path, cv2.IMREAD_GRAYSCALE)
-        print(img_path)
-        print(img_)
         batch_img_lst.append(img_)
         batch_msk_lst.append(msk_)
         bi += 1
         if bi == 4:
             batch_res = eval_batch_scales(eval_net, batch_img_lst, scales=[1.0],
                                           base_crop_size=513, flip=True)
+            print(batch_msk_lst)
             for mi in range(16):
                 hist += cal_hist(batch_msk_lst[mi].flatten(), batch_res[mi].flatten(), 6)