diff --git a/tools/misc/verify_dataset.py b/tools/misc/verify_dataset.py
index 4b5eb7e0e..05e7bb931 100644
--- a/tools/misc/verify_dataset.py
+++ b/tools/misc/verify_dataset.py
@@ -7,7 +7,8 @@ from pathlib import Path
 from mmengine import (Config, DictAction, track_parallel_progress,
                       track_progress)
 
-from mmcls.datasets import PIPELINES, build_dataset
+from mmcls.datasets import build_dataset
+from mmcls.registry import TRANSFORMS
 
 
 def parse_args():
@@ -46,15 +47,14 @@ def parse_args():
 class DatasetValidator():
     """the dataset tool class to check if all file are broken."""
 
-    def __init__(self, dataset_cfg, log_file_path, phase):
+    def __init__(self, dataset_cfg, log_file_path):
         super(DatasetValidator, self).__init__()
         # keep only LoadImageFromFile pipeline
-        assert dataset_cfg.data[phase].pipeline[0][
-            'type'] == 'LoadImageFromFile', 'This tool is only for dataset ' \
-            'that needs to load image from files.'
-        self.pipeline = PIPELINES.build(dataset_cfg.data[phase].pipeline[0])
-        dataset_cfg.data[phase].pipeline = []
-        dataset = build_dataset(dataset_cfg.data[phase])
+        assert dataset_cfg.pipeline[0]['type'] == 'LoadImageFromFile', (
+            'This tool is only for datasets needs to load image from files.')
+        self.pipeline = TRANSFORMS.build(dataset_cfg.pipeline[0])
+        dataset_cfg.pipeline = []
+        dataset = build_dataset(dataset_cfg)
 
         self.dataset = dataset
         self.log_file_path = log_file_path
@@ -102,13 +102,22 @@ def main():
     # touch output file to save broken files list.
     output_path = Path(args.out_path)
     if not output_path.parent.exists():
-        raise Exception('log_file parent directory not found.')
+        raise Exception("Path '--out-path' parent directory not found.")
     if output_path.exists():
         os.remove(output_path)
     output_path.touch()
 
-    # do valid
-    validator = DatasetValidator(cfg, output_path, args.phase)
+    if args.phase == 'train':
+        dataset_cfg = cfg.train_dataloader.dataset
+    elif args.phase == 'val':
+        dataset_cfg = cfg.val_dataloader.dataset
+    elif args.phase == 'test':
+        dataset_cfg = cfg.test_dataloader.dataset
+    else:
+        raise ValueError("'--phase' only support 'train', 'val' and 'test'.")
+
+    # do validate
+    validator = DatasetValidator(dataset_cfg, output_path)
 
     if args.num_process > 1:
         # The default chunksize calcuation method of Pool.map