Skip to content

MetaAnomaly: A Unified Framework for Video Action Recognition.

License

Notifications You must be signed in to change notification settings

IceIce1ce/MetaRecognition

This is the official repository of

MetaRecognition: A Unified Framework for Video Action Recognition.

Setup

conda create -n meta_recognition python=3.10
conda activate meta_recognition
conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 cudatoolkit=11.3 -c pytorch
pip install -r requirements.txt

Download pretrained weights

To train the models on UCF-101 dataset, you can download our pretrained weights from MetaRecognition

Dataset Preparation

Download the UCF-101 dataset and structure the data as follows:

dataset/UCF-101/
  ApplyEyeMakeup
    .avi
  ApplyLipstick
    .avi
  Archery
    .avi
  ...

Usage

To use our model, follow the code snippet below:

cd Video_Recognition

# Train, Test and Demo C3D
bash train_c3d.sh
bash demo_c3d.sh

# Train, Test and Demo DenseNet 3D
bash train_densenet3d.sh
bash demo_densenet3d.sh

# Train, Test and Demo LeNet 3D
bash train_lenet3d.sh
bash demo_lenet3d.sh

# Train, Test and Demo Mobilenet_v1 3D
bash train_mobilenet3d_v1.sh
bash demo_mobilenet3d_v1.sh

# Train, Test and Demo Mobilenet_v2 3D
bash train_mobilenet3d_v2.sh
bash demo_mobilenet3d_v2.sh

# Train, Test and Demo R(2+1)D
bash train_r2plus1d.sh
bash demo_r2plus1d.sh

# Train, Test and Demo R3D
bash train_r3d.sh
bash demo_r3d.sh

# Train, Test and Demo ResNet 3D
bash train_resnet3d.sh
bash demo_resnet3d.sh

# Train, Test and Demo ResNet + I3D
bash train_resnet_i3d.sh
bash demo_resnet_i3d.sh

# Train, Test and Demo ResNeXt 3D
bash train_resnext3d.sh
bash demo_resnext3d.sh

# Train, Test and Demo ShuffleNet_v1 3D
bash train_shufflenet3d_v1.sh
bash demo_shufflenet3d_v1.sh

# Train, Test and Demo ShuffleNet_v2 3D
bash train_shufflenet3d_v2.sh
bash demo_shufflenet3d_v2.sh

# Train, Test and Demo SlowFast
bash train_slowfast.sh
bash demo_slowfast.sh

# Train, Test and Demo SqueezeNet 3D
bash train_squeezenet3d.sh
bash demo_squeezenet3d.sh

# Train, Test and Demo WideResNet 3D
bash train_wideresnet3d.sh
bash demo_wideresnet3d.sh

# Train, Test and Demo Baseline Nonlocal
bash train_baseline_nonlocal.sh
bash demo_baseline_nonlocal.sh

# Train, Test and Demo ARTNet
bash train_artnet.sh
bash demo_artnet.sh

# Train, Test and Demo S3D
bash train_s3d.sh
bash demo_s3d.sh

# Train, Test and Demo I3D
bash train_i3d.sh
bash demo_i3d.sh

# Train, Test and Demo FstCN
bash train_fstcn.sh
bash demo_fstcn.sh

# Train, Test and Demo LRCN
bash train_lrcn.sh
bash demo_lrcn.sh

# Train, Test and Demo FFC_C3D
bash train_ffc_c3d.sh
bash demo_ffc_c3d.sh

# Train, Test and Demo MnasNet
bash train_mnasnet.sh
bash demo_mnasnet.sh

# Train, Test and Demo Baseline_Spectral_norm
bash train_baseline_spectral_norm.sh
bash demo_baseline_spectral_norm.sh

# Train, Test and Demo I3D_pretrained
bash train_i3d_pretrained.sh
bash demo_i3d_pretrained.sh

MetaAnomaly Model Zoo

TBA.

Citation

If you find our work useful, please cite the following:

@misc{Chi2023,
  author       = {Chi Tran},
  title        = {MetaRecognition: A Unified Framework for Video Action Recognition},
  publisher    = {GitHub},
  booktitle    = {GitHub repository},
  howpublished = {https://github.com/IceIce1ce/MetaRecognition},
  year         = {2023}
}

Contact

If you have any questions, feel free to contact Chi Tran ([email protected]).

Acknowledgement

Our framework is built using multiple open source, thanks for their great contributions.

About

MetaAnomaly: A Unified Framework for Video Action Recognition.

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published