README.md 2.1 KB
Newer Older
1 2
# I3D

3 4 5
## Model Zoo

### Kinetics-400
L
linjintao 已提交
6

X
xusu 已提交
7
|config | backbone |pretrain| top1 acc| top5 acc | gpu_mem(M)| ckpt | log|
8
|-|-|-|-|-|-|-|-|
L
linjintao 已提交
9 10 11 12
|[i3d_r34_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_r34_32x2x1_100e_kinetics400_rgb.py) | ResNet34|ImageNet |68.37|88.15|3176| [ckpt]() | [log]()|
|[i3d_r50_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_r50_32x2x1_100e_kinetics400_rgb.py) | ResNet50|ImageNet |72.68|90.78|5170|[ckpt]() | [log]()|
|[i3d_r50_dense_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_r50_dense_32x2x1_100e_kinetics400_rgb.py) | ResNet50| ImageNet|72.77|90.57|5170| [ckpt]() | [log]()|
|[i3d_r50_fast_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_r50_fast_32x2x1_100e_kinetics400_rgb.py) | ResNet50 |ImageNet|72.32|90.72|5170| [ckpt]() | [log]()|
X
xusu 已提交
13
|[i3d_r50_video_3d_32x2x1_100e_kinetics400_rgb](/configs/recognition/i3d/i3d_r50_video_32x2x1_100e_kinetics400_rgb.py) | ResNet50| ImageNet| x | x | x| [ckpt]() | [log]()|
14

L
lixuanyi 已提交
15
For more details on data preparation, you can refer to Kinetics400 in [Data Preparation](/docs/data_preparation.md).
16

17
## Train
18 19 20 21 22
You can use the following command to train a model.
```shell
python tools/train.py ${CONFIG_FILE} [optional arguments]
```

23
Example: train I3D model on Kinetics-400 dataset in a deterministic option with periodic validation.
24 25 26 27 28 29
```shell
python tools/train.py configs/recognition/i3d/i3d_r50_32x2x1_100e_kinetics400_rgb.py \
    --work_dir work_dirs/i3d_r50_32x2x1_100e_kinetics400_rgb \
    --validate --seed 0 --deterministic
```

L
lixuanyi 已提交
30
For more details, you can refer to **Training setting** part in [getting_started](/docs/getting_started.md).
31

32
## Test
33 34 35 36 37
You can use the following command to test a model.
```shell
python tools/test.py ${CONFIG_FILE} ${CHECKPOINT_FILE} [optional arguments]
```

38
Example: test I3D model on Kinetics-400 dataset and dump the result to a json file.
39 40 41 42 43 44
```shell
python tools/test.py configs/recognition/i3d/i3d_r50_32x2x1_100e_kinetics400_rgb.py \
    checkpoints/SOME_CHECKPOINT.pth --eval top_k_accuracy mean_class_accuracy \
    --out result.json
```

L
lixuanyi 已提交
45
For more details, you can refer to **Test a dataset** part in [getting_started](/docs/getting_started.md).