Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Fix] fix flip config of sthsth dataset #2247

Merged
merged 6 commits into from
Apr 6, 2023
Merged

[Fix] fix flip config of sthsth dataset #2247

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
1271914
Add lip_label_map for Flip in some configs
leexinhao Feb 12, 2023
ada1265
update readme
cir7 Feb 21, 2023
e4dae33
fix num_crops in test_pipeline of TSM
cir7 Mar 16, 2023
f8d6c35
update tsm
cir7 Mar 17, 2023
e968666
update tsm-r101
cir7 Mar 20, 2023
bdce110
fix meatfile
cir7 Mar 20, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion configs/recognition/tpn/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ Visual tempo characterizes the dynamics and the temporal scale of an action. Mod

| frame sampling strategy | resolution | gpus | backbone | pretrain | top1 acc | top5 acc | reference top1 acc | reference top5 acc | testing protocol | inference time(video/s) | gpu_mem(M) | config | ckpt | log |
| :---------------------: | :--------: | :--: | :------: | :------: | :------: | :------: | :----------------: | :----------------: | :--------------: | :---------------------: | :--------: | :---------------: | :-------------: | :------------: |
| 1x1x8 | height 100 | 8x6 | ResNet50 | TSM | 48.98 | 78.91 | x | x | 8 clips x 3 crop | x | 8828 | [config](/configs/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb_20220913-d2f5c300.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb.log) |
| 1x1x8 | height 100 | 8x6 | ResNet50 | TSM | 51.87 | 79.67 | x | x | 8 clips x 3 crop | x | 8828 | [config](/configs/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb_20230221-940a3615.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb.log) |

:::{note}

Expand Down
6 changes: 3 additions & 3 deletions configs/recognition/tpn/metafile.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,8 +66,8 @@ Models:
Results:
- Dataset: SthV1
Metrics:
Top 1 Accuracy: 48.98
Top 5 Accuracy: 78.91
Top 1 Accuracy: 51.87
Top 5 Accuracy: 79.67
Task: Action Recognition
Training Log: https://download.openmmlab.com/mmaction/v1.0/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb.log
Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb_20220913-d2f5c300.pth
Weights: (https://download.openmmlab.com/mmaction/v1.0/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb_20230221-940a3615.pth
4 changes: 3 additions & 1 deletion configs/recognition/tpn/tpn-tsm_imagenet-pretrained-r50_8xb8-1x1x8-150e_sthv1-rgb.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,12 +8,14 @@
ann_file_train = 'data/sthv1/sthv1_train_list_rawframes.txt'
ann_file_val = 'data/sthv1/sthv1_val_list_rawframes.txt'
ann_file_test = 'data/sthv1/sthv1_val_list_rawframes.txt'

sthv1_flip_label_map = {2: 4, 4: 2, 30: 41, 41: 30, 52: 66, 66: 52}
train_pipeline = [
dict(type='SampleFrames', clip_len=1, frame_interval=1, num_clips=8),
dict(type='RawFrameDecode'),
dict(type='RandomResizedCrop'),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Flip', flip_ratio=0.5, flip_label_map=sthv1_flip_label_map),
dict(type='ColorJitter'),
dict(type='FormatShape', input_format='NCHW'),
dict(type='PackActionInputs')
Expand Down
10 changes: 5 additions & 5 deletions configs/recognition/tsm/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,11 +32,11 @@ The explosive growth in video streaming gives rise to challenges on performing v

### Something-something V2

| frame sampling strategy | resolution | gpus | backbone | pretrain | top1 acc | top5 acc | testing protocol | FLOPs | params | config | ckpt | log |
| :---------------------: | :--------: | :--: | :-------: | :------: | :------: | :------: | :----------------: | :----: | :----: | :---------------------------------: | :-------------------------------: | :------------------------------: |
| 1x1x8 | 224x224 | 8 | ResNet50 | ImageNet | 60.20 | 86.13 | 8 clips x 10 crop | 32.88G | 23.87M | [config](/configs/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb_20221122-446d261a.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb.log) |
| 1x1x16 | 224x224 | 8 | ResNet50 | ImageNet | 62.46 | 87.75 | 16 clips x 10 crop | 65.75G | 23.87M | [config](/configs/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb_20221122-b1fb8264.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb.log) |
| 1x1x8 | 224x224 | 8 | ResNet101 | ImageNet | 60.49 | 85.99 | 8 clips x 10 crop | 62.66G | 42.86M | [config](/configs/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb_20221122-cb2cc64e.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb.log) |
| frame sampling strategy | resolution | gpus | backbone | pretrain | top1 acc | top5 acc | testing protocol | FLOPs | params | config | ckpt | log |
| :---------------------: | :--------: | :--: | :-------: | :------: | :------: | :------: | :---------------: | :----: | :----: | :---------------------------------: | :-------------------------------: | :-------------------------------: |
| 1x1x8 | 224x224 | 8 | ResNet50 | ImageNet | 62.72 | 87.70 | 8 clips x 3 crop | 32.88G | 23.87M | [config](/configs/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb_20230317-be0fc26e.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb.log) |
| 1x1x16 | 224x224 | 8 | ResNet50 | ImageNet | 64.16 | 88.61 | 16 clips x 3 crop | 65.75G | 23.87M | [config](/configs/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb_20230317-ec6696ad.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb.log) |
| 1x1x8 | 224x224 | 8 | ResNet101 | ImageNet | 63.70 | 88.28 | 8 clips x 3 crop | 62.66G | 42.86M | [config](/configs/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb.py) | [ckpt](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb_20230320-efcc0d1b.pth) | [log](https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb.log) |

1. 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.
2. 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.
Expand Down
34 changes: 17 additions & 17 deletions configs/recognition/tsm/metafile.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -178,60 +178,60 @@ Models:
Parameters: 23.87M
Pretrained: ImageNet
Resolution: 224x224
Training Data: Kinetics-400
Training Data: SthV2
Training Resources: 8 GPUs
Modality: RGB
Results:
- Dataset: Kinetics-400
- Dataset: SthV2
Task: Action Recognition
Metrics:
Top 1 Accuracy: 60.20
Top 5 Accuracy: 86.13
Top 1 Accuracy: 62.72
Top 5 Accuracy: 87.70
Training Log: https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb.log
Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb_20221122-446d261a.pth
Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb_20230317-be0fc26e.pth

- Name: tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb
Config: configs/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb.py
In Collection: TSM
Metadata:
Architecture: ResNet50
Batch Size: 16
Epochs: 100
Epochs: 50
FLOPs: 65.75G
Parameters: 23.87M
Pretrained: ImageNet
Resolution: 224x224
Training Data: Kinetics-400
Training Data: SthV2
Training Resources: 8 GPUs
Modality: RGB
Results:
- Dataset: Kinetics-400
- Dataset: SthV2
Task: Action Recognition
Metrics:
Top 1 Accuracy: 62.46
Top 5 Accuracy: 87.75
Top 1 Accuracy: 64.16
Top 5 Accuracy: 88.61
Training Log: https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb.log
Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb_20221122-b1fb8264.pth
Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb_20230317-ec6696ad.pth

- Name: tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb
Config: configs/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb.py
In Collection: TSM
Metadata:
Architecture: ResNet101
Batch Size: 16
Epochs: 100
Epochs: 50
FLOPs: 62.66G
Parameters: 42.86M
Pretrained: ImageNet
Resolution: 224x224
Training Data: Kinetics-400
Training Data: SthV2
Training Resources: 8 GPUs
Modality: RGB
Results:
- Dataset: Kinetics-400
- Dataset: SthV2
Task: Action Recognition
Metrics:
Top 1 Accuracy: 60.49
Top 5 Accuracy: 85.99
Top 1 Accuracy: 63.70
Top 5 Accuracy: 88.28
Training Log: https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb.log
Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb_20221122-cb2cc64e.pth
Weights: https://download.openmmlab.com/mmaction/v1.0/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb_20230320-efcc0d1b.pth
2 changes: 1 addition & 1 deletion configs/recognition/tsm/tsm_imagenet-pretrained-r101_8xb16-1x1x8-50e_sthv2-rgb.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
_base_ = ['tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb.py']

# model settings
r101_checkpoint = 'https://download.pytorch.org/models/resnet101-cd907fc2.pth'
r101_checkpoint = 'torchvision://resnet101'

model = dict(backbone=dict(pretrained=r101_checkpoint, depth=101))
5 changes: 3 additions & 2 deletions configs/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x16-50e_sthv2-rgb.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@

file_client_args = dict(io_backend='disk')

sthv2_flip_label_map = {86: 87, 87: 86, 93: 94, 94: 93, 166: 167, 167: 166}
train_pipeline = [
dict(type='DecordInit', **file_client_args),
dict(type='SampleFrames', clip_len=1, frame_interval=1, num_clips=16),
Expand All @@ -17,7 +18,7 @@
max_wh_scale_gap=1,
num_fixed_crops=13),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Flip', flip_ratio=0.5, flip_label_map=sthv2_flip_label_map),
dict(type='FormatShape', input_format='NCHW'),
dict(type='PackActionInputs')
]
Expand Down Expand Up @@ -46,7 +47,7 @@
test_mode=True),
dict(type='DecordDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='TenCrop', crop_size=224),
dict(type='ThreeCrop', crop_size=256),
dict(type='FormatShape', input_format='NCHW'),
dict(type='PackActionInputs')
]
Expand Down
5 changes: 3 additions & 2 deletions configs/recognition/tsm/tsm_imagenet-pretrained-r50_8xb16-1x1x8-50e_sthv2-rgb.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,7 @@

file_client_args = dict(io_backend='disk')

sthv2_flip_label_map = {86: 87, 87: 86, 93: 94, 94: 93, 166: 167, 167: 166}
train_pipeline = [
dict(type='DecordInit', **file_client_args),
dict(type='SampleFrames', clip_len=1, frame_interval=1, num_clips=8),
Expand All @@ -24,7 +25,7 @@
max_wh_scale_gap=1,
num_fixed_crops=13),
dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Flip', flip_ratio=0.5, flip_label_map=sthv2_flip_label_map),
dict(type='FormatShape', input_format='NCHW'),
dict(type='PackActionInputs')
]
Expand Down Expand Up @@ -53,7 +54,7 @@
twice_sample=True),
dict(type='DecordDecode'),
dict(type='Resize', scale=(-1, 256)),
dict(type='TenCrop', crop_size=224),
dict(type='ThreeCrop', crop_size=256),
dict(type='FormatShape', input_format='NCHW'),
dict(type='PackActionInputs')
]
Expand Down
Loading