diff --git a/mmyolo/utils/misc.py b/mmyolo/utils/misc.py
index c90f52b94..f0e4d3863 100644
--- a/mmyolo/utils/misc.py
+++ b/mmyolo/utils/misc.py
@@ -1,9 +1,12 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import os
 import urllib
+from copy import deepcopy
+from typing import List, Union
 
 import numpy as np
 import torch
+from mmengine.config import ConfigDict
 from mmengine.utils import scandir
 from prettytable import PrettyTable
 
@@ -55,7 +58,7 @@ def auto_arrange_images(image_list: list, image_column: int = 2) -> np.ndarray:
     return image_show
 
 
-def get_file_list(source_root: str) -> [list, dict]:
+def get_file_list(source_root: str) -> Union[list, dict]:
     """Get file list.
 
     Args:
@@ -131,3 +134,33 @@ def judge_keys(dataloader_cfg):
     judge_keys(cfg.get('train_dataloader', {}))
     judge_keys(cfg.get('val_dataloader', {}))
     judge_keys(cfg.get('test_dataloader', {}))
+
+
+def get_cfg_class_name(cfg_type: str):
+    return cfg_type.split('.')[-1]
+
+
+def convert_to_val_pipeline(data_cfg: ConfigDict,
+                            val_pipeline: List) -> ConfigDict:
+
+    if get_cfg_class_name(data_cfg.get('type')) in [
+            'MultiImageMixDataset', 'ClassBalancedDataset'
+    ]:
+        # While evaluation, there shouldn't use multi image or oversample.
+        data_cfg = deepcopy(data_cfg.get('dataset'))
+        data_cfg['pipeline'] = val_pipeline
+    elif 'dataset' in data_cfg:
+        data = deepcopy(data_cfg.get('dataset'))
+        if isinstance(data, list):
+            # concat dataset
+            data_cfg['dataset'] = [
+                convert_to_val_pipeline(d, val_pipeline) for d in data
+            ]
+        else:
+            # Deal with other dataset wrapper
+            data_cfg['dataset'] = convert_to_val_pipeline(data, val_pipeline)
+    else:
+        # Regular dataset
+        data_cfg['pipeline'] = val_pipeline
+    data_cfg['test_mode'] = True
+    return data_cfg
diff --git a/tools/test.py b/tools/test.py
index c05defe3c..91f499e5b 100644
--- a/tools/test.py
+++ b/tools/test.py
@@ -2,6 +2,7 @@
 import argparse
 import os
 import os.path as osp
+import warnings
 
 from mmdet.engine.hooks.utils import trigger_visualization_hook
 from mmengine.config import Config, ConfigDict, DictAction
@@ -10,6 +11,7 @@
 
 from mmyolo.registry import RUNNERS
 from mmyolo.utils import is_metainfo_lower
+from mmyolo.utils.misc import convert_to_val_pipeline
 
 
 # TODO: support fuse_conv_bn
@@ -35,6 +37,13 @@ def parse_args():
         '--tta',
         action='store_true',
         help='Whether to use test time augmentation')
+    parser.add_argument(
+        '--phase',
+        default='test',
+        type=str,
+        choices=['train', 'test', 'val'],
+        help='phase of dataset to test, accept "train" "test" and "val". '
+        'Defaults to "test".')
     parser.add_argument(
         '--show', action='store_true', help='show prediction results')
     parser.add_argument(
@@ -109,6 +118,29 @@ def main():
     # Determine whether the custom metainfo fields are all lowercase
     is_metainfo_lower(cfg)
 
+    #
+    if args.phase == 'train':
+        # If test on train phase, it will use val pipline
+        val_data_cfg = cfg.val_dataloader.dataset
+        while 'dataset' in val_data_cfg:
+            val_data_cfg = val_data_cfg['dataset']
+        val_pipeline = val_data_cfg.pipeline
+        train_dataset = cfg.train_dataloader.dataset
+        test_data_cfg = convert_to_val_pipeline(train_dataset, val_pipeline)
+        cfg.test_dataloader.dataset = test_data_cfg
+        evaluator_cfg = cfg.val_evaluator
+        if evaluator_cfg.get('ann_file') is not None:
+            evaluator_cfg.pop('ann_file')
+            warnings.warn('When use train phase for test, `ann_file` will '
+                          'be removed to use loaded annotation directly.')
+        cfg.test_evaluator = cfg.val_evaluator
+    elif args.phase == 'val':
+        test_data_cfg = cfg.val_dataloader.dataset
+        cfg.test_dataloader.dataset = cfg.val_dataloader.dataset
+        cfg.test_evaluator = cfg.val_evaluator
+    elif args.phase == 'test':
+        test_data_cfg = cfg.test_dataloader.dataset
+
     if args.tta:
         assert 'tta_model' in cfg, 'Cannot find ``tta_model`` in config.' \
                                    " Can't use tta !"
@@ -116,7 +148,7 @@ def main():
                                       "in config. Can't use tta !"
 
         cfg.model = ConfigDict(**cfg.tta_model, module=cfg.model)
-        test_data_cfg = cfg.test_dataloader.dataset
+        # test_data_cfg = cfg.test_dataloader.dataset
         while 'dataset' in test_data_cfg:
             test_data_cfg = test_data_cfg['dataset']