open-mmlab · ly015 · Dec 22, 2022 · Dec 6, 2022 · Dec 6, 2022 · Dec 7, 2022
diff --git a/configs/recognition/c2d/README.md b/configs/recognition/c2d/README.md
@@ -21,18 +21,18 @@ Both convolutional and recurrent operations are building blocks that process one
 
 ### Kinetics-400
 
-| frame sampling strategy | scheduler |   resolution   | gpus |   backbone    | pretrain | top1 acc | top5 acc |   reference top1 acc    |   reference top5 acc    |  testing protocol  | gpu_mem(M) | FLOPs | params |   config    |   ckpt    |   log    |
-| :---------------------: | :-------: | :------------: | :--: | :-----------: | :------: | :------: | :------: | :---------------------: | :---------------------: | :----------------: | :--------: | :---: | :----: | :---------: | :-------: | :------: |
-|          8x8x1          | MultiStep | short-side 320 |  8   | ResNet50<br>  | ImageNet |  73.16   |  90.88   | 67.2<br>[\[PySlowFast\]](https://github.com/facebookresearch/SlowFast/blob/main/MODEL_ZOO.md#kinetics-400-and-600) | 87.8<br>[\[PySlowFast\]](https://github.com/facebookresearch/SlowFast/blob/main/MODEL_ZOO.md#kinetics-400-and-600) | 10 clips x 3 crops |   21547    |  33G  | 24.3M  | [config](/configs/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb_20221027-e0227b22.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.log) |
-|          8x8x1          | MultiStep | short-side 320 |  8   | ResNet101<br> | ImageNet |  74.57   |  91.60   |            x            |            x            | 10 clips x 3 crops |   31836    |  63G  | 43.3M  | [config](/configs/recognition/c2d/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb_20221027-557bd8bc.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.log) |
-|          8x8x1          | MultiStep | short-side 320 |  8   | ResNet50<br>(TemporalPool) | ImageNet |  73.57   |  90.96   | 71.9<br>[\[Non-Local\]](https://github.com/facebookresearch/video-nonlocal-net#modifications-for-improving-speed) | 90.0<br>[\[Non-Local\]](https://github.com/facebookresearch/video-nonlocal-net#modifications-for-improving-speed) | 10 clips x 3 crops |   17006    |  19G  | 24.3M  | [config](/configs/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb_20221027-3ca304fa.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb.log) |
-|         16x4x1          | MultiStep | short-side 320 |  8   | ResNet50<br>(TemporalPool) | ImageNet |  74.54   |  91.76   |            x            |            x            | 10 clips x 3 crops |   33630    |  39G  | 24.3M  | [config](/configs/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb_20221027-5f382a43.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb.log) |
+| frame sampling strategy | scheduler | resolution | gpus |    backbone     | pretrain | top1 acc | top5 acc |    reference top1 acc     |    reference top5 acc     | testing protocol  | FLOPs | params |     config     |     ckpt     |     log     |
+| :---------------------: | :-------: | :--------: | :--: | :-------------: | :------: | :------: | :------: | :-----------------------: | :-----------------------: | :---------------: | :---: | :----: | :------------: | :----------: | :---------: |
+|          8x8x1          | MultiStep |  224x224   |  8   |  ResNet50<br>   | ImageNet |  73.44   |  91.00   | 67.2<br>[\[PySlowFast\]](https://github.com/facebookresearch/SlowFast/blob/main/MODEL_ZOO.md#kinetics-400-and-600) | 87.8<br>[\[PySlowFast\]](https://github.com/facebookresearch/SlowFast/blob/main/MODEL_ZOO.md#kinetics-400-and-600) | 10 clips x 3 crop |  33G  | 24.3M  | [config](/configs/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb_20221027-e0227b22.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.log) |
+|          8x8x1          | MultiStep |  224x224   |  8   |  ResNet101<br>  | ImageNet |  74.97   |  91.77   |             x             |             x             | 10 clips x 3 crop |  63G  | 43.3M  | [config](/configs/recognition/c2d/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb_20221027-557bd8bc.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.log) |
+|          8x8x1          | MultiStep |  224x224   |  8   | ResNet50<br>(TemporalPool) | ImageNet |  73.89   |  91.21   | 71.9<br>[\[Non-Local\]](https://github.com/facebookresearch/video-nonlocal-net#modifications-for-improving-speed) | 90.0<br>[\[Non-Local\]](https://github.com/facebookresearch/video-nonlocal-net#modifications-for-improving-speed) | 10 clips x 3 crop |  19G  | 24.3M  | [config](/configs/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb_20221027-3ca304fa.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb.log) |
+|         16x4x1          | MultiStep |  224x224   |  8   | ResNet50<br>(TemporalPool) | ImageNet |  74.97   |  91.91   |             x             |             x             | 10 clips x 3 crop |  39G  | 24.3M  | [config](/configs/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb_20221027-5f382a43.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb.log) |
 
 1. The values in columns named after "reference" are the results reported in the original repo, using the same model settings.
 2. The **gpus** indicates the number of gpus we used to get the checkpoint. If you want to use a different number of gpus or videos per gpu, the best way is to set `--auto-scale-lr` when calling `tools/train.py`, this parameter will auto-scale the learning rate according to the actual batch size and the original batch size.
 3. The validation set of Kinetics400 we used consists of 19796 videos. These videos are available at [Kinetics400-Validation](https://mycuhk-my.sharepoint.com/:u:/g/personal/1155136485_link_cuhk_edu_hk/EbXw2WX94J1Hunyt3MWNDJUBz-nHvQYhO9pvKqm6g39PMA?e=a9QldB). The corresponding [data list](https://download.openmmlab.com/mmaction/dataset/k400_val/kinetics_val_list.txt) (each line is of the format 'video_id, num_frames, label_index') and the [label map](https://download.openmmlab.com/mmaction/dataset/k400_val/kinetics_class2ind.txt) are also available.
 
-For more details on data preparation, you can refer to [preparing_kinetics](/tools/data/kinetics/README.md).
+For more details on data preparation, you can refer to [Kinetics400](/tools/data/kinetics/README.md).
 
 ## Train
 
@@ -46,7 +46,7 @@ Example: train C2D model on Kinetics-400 dataset in a deterministic option with
 
 ```shell
 python tools/train.py configs/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb.py  \
-    --cfg-options randomness.seed=0 randomness.deterministic=True
+    --seed 0 --deterministic
 ```
 
 For more details, you can refer to the **Training** part in the [Training and Test Tutorial](/docs/en/user_guides/4_train_test.md).

diff --git a/configs/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.py b/configs/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.py
@@ -10,12 +10,7 @@
 ann_file_train = 'data/kinetics400/kinetics400_train_list_videos.txt'
 ann_file_val = 'data/kinetics400/kinetics400_val_list_videos.txt'
 
-# file_client_args = dict(
-#     io_backend='petrel',
-#     path_mapping=dict(
-#         {'data/kinetics400': 's3://openmmlab/datasets/action/Kinetics400'}))
 file_client_args = dict(io_backend='disk')
-
 train_pipeline = [
     dict(type='DecordInit', **file_client_args),
     dict(type='SampleFrames', clip_len=8, frame_interval=8, num_clips=1),

diff --git a/configs/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb.py b/configs/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb.py
@@ -10,12 +10,7 @@
 ann_file_train = 'data/kinetics400/kinetics400_train_list_videos.txt'
 ann_file_val = 'data/kinetics400/kinetics400_val_list_videos.txt'
 
-# file_client_args = dict(
-#     io_backend='petrel',
-#     path_mapping=dict(
-#         {'data/kinetics400': 's3://openmmlab/datasets/action/Kinetics400'}))
 file_client_args = dict(io_backend='disk')
-
 train_pipeline = [
     dict(type='DecordInit', **file_client_args),
     dict(type='SampleFrames', clip_len=16, frame_interval=4, num_clips=1),

diff --git a/configs/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb.py b/configs/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb.py
@@ -10,12 +10,7 @@
 ann_file_train = 'data/kinetics400/kinetics400_train_list_videos.txt'
 ann_file_val = 'data/kinetics400/kinetics400_val_list_videos.txt'
 
-# file_client_args = dict(
-#     io_backend='petrel',
-#     path_mapping=dict(
-#         {'data/kinetics400': 's3://openmmlab/datasets/action/Kinetics400'}))
 file_client_args = dict(io_backend='disk')
-
 train_pipeline = [
     dict(type='DecordInit', **file_client_args),
     dict(type='SampleFrames', clip_len=8, frame_interval=8, num_clips=1),

diff --git a/configs/recognition/c2d/metafile.yml b/configs/recognition/c2d/metafile.yml
@@ -16,16 +16,16 @@ Models:
       FLOPs: 33G
       Parameters: 24.3M
       Pretrained: ImageNet
-      Resolution: short-side 320
+      Resolution: 224x224
       Training Data: Kinetics-400
       Training Resources: 8 GPUs
     Modality: RGB
     Results:
       - Dataset: Kinetics-400
         Task: Action Recognition
         Metrics:
-              Top 1 Accuracy: 73.16
-              Top 5 Accuracy: 90.88
+              Top 1 Accuracy: 73.44
+              Top 5 Accuracy: 91.00
     Training Log: https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.log
     Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb_20221027-e0227b22.pth
 
@@ -39,16 +39,16 @@ Models:
       FLOPs: 63G
       Parameters: 43.3M
       Pretrained: ImageNet
-      Resolution: short-side 320
+      Resolution: 224x224
       Training Data: Kinetics-400
       Training Resources: 8 GPUs
     Modality: RGB
     Results:
     - Dataset: Kinetics-400
       Task: Action Recognition
       Metrics:
-        Top 1 Accuracy: 74.57
-        Top 5 Accuracy: 91.60
+        Top 1 Accuracy: 74.97
+        Top 5 Accuracy: 91.77
     Training Log: https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb.log
     Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r101-in1k-pre-nopool_8xb32-8x8x1-100e_kinetics400-rgb_20221027-557bd8bc.pth
 
@@ -62,16 +62,16 @@ Models:
       FLOPs: 19G
       Parameters: 24.3M
       Pretrained: ImageNet
-      Resolution: short-side 320
+      Resolution: 224x224
       Training Data: Kinetics-400
       Training Resources: 8 GPUs
     Modality: RGB
     Results:
     - Dataset: Kinetics-400
       Task: Action Recognition
       Metrics:
-        Top 1 Accuracy: 73.57
-        Top 5 Accuracy: 90.96
+        Top 1 Accuracy: 73.89
+        Top 5 Accuracy: 91.21
     Training Log: https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb.log
     Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-8x8x1-100e_kinetics400-rgb_20221027-3ca304fa.pth
 
@@ -85,15 +85,15 @@ Models:
       FLOPs: 39G
       Parameters: 24.3M
       Pretrained: ImageNet
-      Resolution: short-side 320
+      Resolution: 224x224
       Training Data: Kinetics-400
       Training Resources: 8 GPUs
     Modality: RGB
     Results:
     - Dataset: Kinetics-400
       Task: Action Recognition
       Metrics:
-        Top 1 Accuracy: 74.54
-        Top 5 Accuracy: 91.76
+        Top 1 Accuracy: 74.97
+        Top 5 Accuracy: 91.91
     Training Log: https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb.log
     Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/c2d/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb/c2d_r50-in1k-pre_8xb32-16x4x1-100e_kinetics400-rgb_20221027-5f382a43.pth
diff --git a/configs/recognition/c3d/README.md b/configs/recognition/c3d/README.md
@@ -20,16 +20,14 @@ We propose a simple, yet effective approach for spatiotemporal feature learning
 
 ### UCF-101
 
-| frame sampling strategy | resolution | gpus | backbone | pretrain | top1 acc | top5 acc | testing protocol  | inference_time(video/s) | gpu_mem(M) |            config            |            ckpt             |            log             |
-| :---------------------: | :--------: | :--: | :------: | :------: | :------: | :------: | :---------------: | :---------------------: | :--------: | :--------------------------: | :-------------------------: | :------------------------: |
-|         16x1x1          |    raw     |  8   |   c3d    | sports1m |  82.92   |  96.11   | 10 clips x 1 crop |            x            |    6067    | [config](/configs/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb_20220811-31723200.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb.log) |
+| frame sampling strategy | resolution | gpus | backbone | pretrain | top1 acc | top5 acc | testing protocol  | FLOPs | params |                config                |                ckpt                |                log                |
+| :---------------------: | :--------: | :--: | :------: | :------: | :------: | :------: | :---------------: | :---: | :----: | :----------------------------------: | :--------------------------------: | :-------------------------------: |
+|         16x1x1          |  112x112   |  8   |   c3d    | sports1m |  83.08   |  95.93   | 10 clips x 1 crop | 38.5G | 78.4M  | [config](/configs/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb_20220811-31723200.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb.log) |
 
 1. The author of C3D normalized UCF-101 with volume mean and used SVM to classify videos, while we normalized the dataset with RGB mean value and used a linear classifier.
-2. The **gpus** indicates the number of gpu (80G A100) we used to get the checkpoint. It is noteworthy that the configs we provide are used for 8 gpus as default.
-   According to the [Linear Scaling Rule](https://arxiv.org/abs/1706.02677), you may set the learning rate proportional to the batch size if you use different GPUs or videos per GPU,
-   e.g., lr=0.01 for 4 GPUs x 2 video/gpu and lr=0.08 for 16 GPUs x 4 video/gpu.
+2. The **gpus** indicates the number of gpus we used to get the checkpoint. If you want to use a different number of gpus or videos per gpu, the best way is to set `--auto-scale-lr` when calling `tools/train.py`, this parameter will auto-scale the learning rate according to the actual batch size and the original batch size.
 
-For more details on data preparation, you can refer to the **Prepare videos** part in the [Data Preparation Tutorial](/docs/en/user_guides/2_data_prepare.md).
+For more details on data preparation, you can refer to [UCF101](/tools/data/ucf101/README.md).
 
 ## Train
 
@@ -43,7 +41,7 @@ Example: train C3D model on UCF-101 dataset in a deterministic option with perio
 
 ```shell
 python tools/train.py configs/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb.py \
-    --cfg-options randomness.seed=0 randomness.deterministic=True
+    --seed=0 --deterministic
 ```
 
 For more details, you can refer to the **Training** part in the [Training and Test Tutorial](/docs/en/user_guides/4_train_test.md).

diff --git a/configs/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb.py b/configs/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb.py
@@ -11,8 +11,10 @@
 ann_file_train = f'data/ucf101/ucf101_train_split_{split}_videos.txt'
 ann_file_val = f'data/ucf101/ucf101_val_split_{split}_videos.txt'
 ann_file_test = f'data/ucf101/ucf101_val_split_{split}_videos.txt'
+
+file_client_args = dict(io_backend='disk')
 train_pipeline = [
-    dict(type='DecordInit'),
+    dict(type='DecordInit', **file_client_args),
     dict(type='SampleFrames', clip_len=16, frame_interval=1, num_clips=1),
     dict(type='DecordDecode'),
     dict(type='Resize', scale=(-1, 128)),
@@ -22,7 +24,7 @@
     dict(type='PackActionInputs')
 ]
 val_pipeline = [
-    dict(type='DecordInit'),
+    dict(type='DecordInit', **file_client_args),
     dict(
         type='SampleFrames',
         clip_len=16,
@@ -36,7 +38,7 @@
     dict(type='PackActionInputs')
 ]
 test_pipeline = [
-    dict(type='DecordInit'),
+    dict(type='DecordInit', **file_client_args),
     dict(
         type='SampleFrames',
         clip_len=16,
@@ -106,3 +108,9 @@
     clip_grad=dict(max_norm=40, norm_type=2))
 
 default_hooks = dict(checkpoint=dict(interval=5))
+
+# Default setting for scaling LR automatically
+#   - `enable` means enable scaling LR automatically
+#       or not by default.
+#   - `base_batch_size` = (8 GPUs) x (30 samples per GPU).
+auto_scale_lr = dict(enable=False, base_batch_size=240)
diff --git a/configs/recognition/c3d/metafile.yml b/configs/recognition/c3d/metafile.yml
@@ -13,17 +13,18 @@ Models:
       Architecture: c3d
       Batch Size: 30
       Epochs: 45
-      FLOPs: 38615475200
-      Parameters: 78409573
+      FLOPs: 38.5G
+      Parameters: 78.4M
       Pretrained: sports1m
+      Resolution: 112x112
       Training Data: UCF101
       Training Resources: 8 GPUs
     Modality: RGB
     Results:
     - Dataset: UCF101
       Task: Action Recognition
       Metrics:
-        Top 1 Accuracy: 82.92
-        Top 5 Accuracy: 96.11
+        Top 1 Accuracy: 83.08
+        Top 5 Accuracy: 95.93
     Training Log: https://download.openmmlab.com/mmaction/v1.0/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb.log
     Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/c3d/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb/c3d_sports1m-pretrained_8xb30-16x1x1-45e_ucf101-rgb_20220811-31723200.pth