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

Jump to bottom

Update performance of video models #256

Merged
merged 1 commit into from
Oct 16, 2020
Merged

Update performance of video models #256

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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/slowonly/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@
|config | resolution | gpus | backbone |pretrain| top1 acc| top5 acc | inference_time(video/s) | gpu_mem(M) | ckpt | log| json|
|:--|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
|[slowonly_r50_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py)|short-side 256|8x4| ResNet50 | None |72.76|90.51|x|3168|[ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb_20200820-bea7701f.pth)|[log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log)|[json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_4x16x1_256e_kinetics400_rgb/20200817_001411.log.json)|
|[slowonly_r50_video_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_video_4x16x1_256e_kinetics400_rgb.py)|short-side 256|8| ResNet50 | None |72.11|90.32|x|3168|[ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_4x16x1_256e_kinetics400_rgb_20200826-f39b8cd8.pth)|[log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_4x16x1_256e_kinetics400_rgb/20200812_172302.log)|[json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_4x16x1_256e_kinetics400_rgb/20200812_172302.log.json)|
|[slowonly_r50_video_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_video_4x16x1_256e_kinetics400_rgb.py)|short-side 320|8x2| ResNet50 | None |72.90|90.82|x|8472|[ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014-c9cdc656.pth)|[log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014.log)|[json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb/slowonly_r50_video_320p_4x16x1_256e_kinetics400_rgb_20201014.json)|
|[slowonly_r50_8x8x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb.py) |short-side 256|8x4| ResNet50 | None |74.42|91.49|x|5820|[ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb_20200820-75851a7d.pth)|[log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/20200817_003320.log)|[json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_256p_8x8x1_256e_kinetics400_rgb/20200817_003320.log.json)|
|[slowonly_r50_4x16x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb.py)|short-side 320|8x2| ResNet50 | None |73.02|90.77|4.0 (40x3 frames)|3168|[ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16x1_256e_kinetics400_rgb_20200704-a69556c6.pth)| [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/so_4x16.log)| [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_4x16x1_256e_kinetics400_rgb/slowonly_r50_4x16_73.02_90.77.log.json)|
|[slowonly_r50_8x8x1_256e_kinetics400_rgb](/configs/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb.py) |short-side 320|8x3| ResNet50 | None |74.93|91.92|2.3 (80x3 frames)|5820| [ckpt](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/slowonly_r50_8x8x1_256e_kinetics400_rgb_20200703-a79c555a.pth) | [log](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/so_8x8.log)| [json](https://download.openmmlab.com/mmaction/recognition/slowonly/slowonly_r50_8x8x1_256e_kinetics400_rgb/slowonly_r50_8x8_74.93_91.92.log.json)|
Expand Down
4 changes: 2 additions & 2 deletions configs/recognition/slowonly/slowonly_r50_video_4x16x1_256e_kinetics400_rgb.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,7 +74,7 @@
dict(type='ToTensor', keys=['imgs'])
]
data = dict(
videos_per_gpu=8,
videos_per_gpu=24,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

v100?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nope, slowfast_4x16 doesn't consume lots of memory, we can fit 24 samples onto a 1080 Ti.

workers_per_gpu=4,
train=dict(
type=dataset_type,
Expand All @@ -93,7 +93,7 @@
pipeline=test_pipeline))
# optimizer
optimizer = dict(
type='SGD', lr=0.1, momentum=0.9,
type='SGD', lr=0.3, momentum=0.9,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

double check that this lr is for 8 gpus

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, that is for 8 gpu (since we use batch size 24 on each GPU

weight_decay=0.0001) # this lr is used for 8 gpus
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
Expand Down
1 change: 1 addition & 0 deletions configs/recognition/tsn/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,7 @@
|[tsn_r50_320p_1x1x8_100e_kinetics400_rgb](/configs/recognition/tsn/tsn_r50_320p_1x1x8_100e_kinetics400_rgb.py) |short-side 320|8x3| ResNet50| ImageNet |72.41|90.55|x|x|11.1 (25x3 frames)| 8344 | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_320p_1x1x8_100e_kinetics400_rgb/tsn_r50_320p_1x1x8_100e_kinetics400_rgb_20200702-ef80e3d7.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_320p_1x1x8_100e_kinetics400_rgb/tsn_r50_f8_kinetics400_shortedge_72.4_90.6.log) | [json](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_320p_1x1x8_100e_kinetics400_rgb/tsn_r50_f8_kinetics400_shortedge_72.4_90.6.log.json)|
|[tsn_r50_320p_1x1x8_110e_kinetics400_flow](/configs/recognition/tsn/tsn_r50_320p_1x1x8_110e_kinetics400_flow.py) |short-side 320|8x4| ResNet50 | ImageNet|57.76|80.99|x|x|x| 8473 | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_320p_1x1x8_110e_kinetics400_flow/tsn_r50_320p_1x1x8_110e_kinetics400_flow_20200705-1f39486b.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_320p_1x1x8_110e_kinetics400_flow/tsn_r50_f8_kinetics400_flow_shortedge_57.8_81.0.log) | [json](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_320p_1x1x8_110e_kinetics400_flow/tsn_r50_f8_kinetics400_flow_shortedge_57.8_81.0.log.json)|
|tsn_r50_320p_1x1x8_kinetics400_twostream [1: 1]* |x|x| ResNet50| ImageNet |74.64|91.77| x | x | x | x | x|x|x|
|[tsn_r50_video_320p_1x1x3_100e_kinetics400_rgb](/configs/recognition/tsn/tsn_r50_video_320p_1x1x3_100e_kinetics400_rgb.py) |short-side 320|8| ResNet50 | ImageNet |71.11|90.04| x | x | x | 8343 | [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_video_320p_1x1x3_100e_kinetics400_rgb/tsn_r50_video_320p_1x1x3_100e_kinetics400_rgb_20201014-5ae1ee79.pth) |[log](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_video_320p_1x1x3_100e_kinetics400_rgb/tsn_r50_video_320p_1x1x3_100e_kinetics400_rgb_20201014.log)|[json](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_video_320p_1x1x3_100e_kinetics400_rgb/tsn_r50_video_320p_1x1x3_100e_kinetics400_rgb_20201014.json)|
|[tsn_r50_dense_1x1x8_100e_kinetics400_rgb](/configs/recognition/tsn/tsn_r50_dense_1x1x8_100e_kinetics400_rgb.py) |340x256|8| ResNet50 | ImageNet|70.77|89.3|[68.75](https://github.com/mit-han-lab/temporal-shift-module/tree/8d53d6fda40bea2f1b37a6095279c4b454d672bd#training)|[88.42](https://github.com/mit-han-lab/temporal-shift-module/tree/8d53d6fda40bea2f1b37a6095279c4b454d672bd#training)|12.2 (8x10 frames)|8344| [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_dense_1x1x8_100e_kinetics400_rgb/tsn_r50_dense_1x1x8_100e_kinetics400_rgb_20200606-e925e6e3.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_dense_1x1x8_100e_kinetics400_rgb/20200606_003901.log)| [json](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_dense_1x1x8_100e_kinetics400_rgb/20200606_003901.log.json)|
|[tsn_r50_video_1x1x8_100e_kinetics400_rgb](/configs/recognition/tsn/tsn_r50_video_1x1x8_100e_kinetics400_rgb.py) |short-side 256|8| ResNet50| ImageNet | 71.79 | 90.25 |x|x|x|21558| [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_video_1x1x8_100e_kinetics400_rgb/tsn_r50_video_1x1x8_100e_kinetics400_rgb_20200702-568cde33.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_video_1x1x8_100e_kinetics400_rgb/tsn_r50_video_2d_1x1x8_100e_kinetics400_rgb.log)| [json](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_video_1x1x8_100e_kinetics400_rgb/tsn_r50_video_2d_1x1x8_100e_kinetics400_rgb.log.json)|
|[tsn_r50_video_dense_1x1x8_100e_kinetics400_rgb](/configs/recognition/tsn/tsn_r50_video_dense_1x1x8_100e_kinetics400_rgb.py) |short-side 256|8| ResNet50| ImageNet | 70.40 | 89.12 |x|x|x|21553| [ckpt](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_video_dense_1x1x8_100e_kinetics400_rgb/tsn_r50_video_dense_1x1x8_100e_kinetics400_rgb_20200703-0f19175f.pth) | [log](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_video_dense_1x1x8_100e_kinetics400_rgb/tsn_r50_video_2d_1x1x8_dense_100e_kinetics400_rgb.log)| [json](https://download.openmmlab.com/mmaction/recognition/tsn/tsn_r50_video_dense_1x1x8_100e_kinetics400_rgb/tsn_r50_video_2d_1x1x8_dense_100e_kinetics400_rgb.log.json)|
Expand Down
7 changes: 1 addition & 6 deletions ...n_r50_video_1x1x3_100e_kinetics400_rgb.py → ..._video_320p_1x1x3_100e_kinetics400_rgb.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,12 +35,7 @@
num_clips=3,
start_index=0),
dict(type='DecordDecode'),
dict(
type='MultiScaleCrop',
input_size=224,
scales=(1, 0.875, 0.75, 0.66),
random_crop=False,
max_wh_scale_gap=1),
dict(type='RandomResizedCrop'),
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

does it mean that RandomResizedCrop is better than MultiScaleCrop in this case?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sure, we already validated that RandomResizedCrop can outperform MultiScaleCrop. The contribution of that PR is to show that training with videos doesn't lead to any performance drop.
Before renaming, this config is named tsn_r50_video_1x1x3_100e_kinetics400_rgb.py, and there is no performance score for that config in README.

dict(type='Resize', scale=(224, 224), keep_ratio=False),
dict(type='Flip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
Expand Down