fix small error

configs/quick_start/ResNet50_vd_multilabel.yam → configs/quick_start/ResNet50_vd_multilabel.yaml

@@ -34,8 +34,8 @@ OPTIMIZER:
 TRAIN:
     batch_size: 256
     num_workers: 4
-    file_list: "./dataset/NUS-SCENE-dataset/multilabel_train_list.txt"
-    data_dir: "./dataset/NUS-SCENE-dataset/images"
+    file_list: "./dataset/NUS-WIDE-SCENE/NUS-SCENE-dataset/multilabel_train_list.txt"
+    data_dir: "./dataset/NUS-WIDE-SCENE/NUS-SCENE-dataset/images"
     shuffle_seed: 0
     transforms:
         - DecodeImage:
@@ -59,8 +59,8 @@ TRAIN:
 VALID:
     batch_size: 64
     num_workers: 4
-    file_list: "./dataset/NUS-SCENE-dataset/multilabel_test_list.txt"
-    data_dir: "./dataset/NUS-SCENE-dataset/images"
+    file_list: "./dataset/NUS-WIDE-SCENE/NUS-SCENE-dataset/multilabel_test_list.txt"
+    data_dir: "./dataset/NUS-WIDE-SCENE/NUS-SCENE-dataset/images"
     shuffle_seed: 0
     transforms:
         - DecodeImage:

ppcls/modeling/loss.py

@@ -89,8 +89,8 @@ class MultiLabelLoss(Loss):
     def __init__(self, class_dim=1000, epsilon=None):
         super(MultiLabelLoss, self).__init__(class_dim, epsilon)
 
-    def __call__(self, input, target, use_pure_fp16=False):
-        cost = self._binary_crossentropy(input, target, use_pure_fp16)
+    def __call__(self, input, target):
+        cost = self._binary_crossentropy(input, target)
 
         return cost
 

tools/infer/infer.py

@@ -72,10 +72,15 @@ def main():
             for number, result_dict in enumerate(batch_result_list):
                 filename = img_path_list[number].split("/")[-1]
                 clas_ids = result_dict["clas_ids"]
-                scores_str = "[{}]".format(", ".join("{:.2f}".format(
-                    r) for r in result_dict["scores"]))
-                print("File:{}, Top-{} result: class id(s): {}, score(s): {}".
-                      format(filename, args.top_k, clas_ids, scores_str))
+                if multilabel:
+                    print("File:{}, multilabel result: ".format(filename))
+                    for id, score in zip(clas_ids, result_dict["scores"]):
+                        print("\tclass id: {}, probability: {:.2f}".format(id, score))
+                else:
+                    scores_str = "[{}]".format(", ".join("{:.2f}".format(
+                        r) for r in result_dict["scores"]))
+                    print("File:{}, Top-{} result: class id(s): {}, score(s): {}".
+                        format(filename, args.top_k, clas_ids, scores_str))
 
                 if args.pre_label_image:
                     save_prelabel_results(clas_ids[0], img_path_list[number],