Skip to content
/ astnet Public

This is an official implementation for "Attention-based Residual Autoencoder for Video Anomaly Detection".

License

Notifications You must be signed in to change notification settings

vt-le/astnet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

79 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ASTNet: Attention-based Residual Autoencoder for Video Anomaly Detection

This is the official implementation of Attention-based Residual Autoencoder for Video Anomaly Detection Hits.

Related works

HSTforU: See HSTforU: Anomaly Detection in Aerial and Ground-based Videos with Hierarchical Spatio-Temporal Transformer for U-net .

MoGuP: See MoGuP:Motion-guided Prediction for Video Anomaly Detection.

Updates

  • [6/01/2023] Training script of ASTNet is released.
  • [5/25/2022] ASTNet is available online.
  • [4/21/2022] Code of ASTNet is released!

Star History

Star History Chart

Prerequisites

  • Linux or macOS
  • Python 3
  • PyTorch 1.7.0

Setup

The code can be run with Python 3.6 and above.

Install the required packages:

pip install -r requirements.txt

Clone this repo:

git clone https://github.com/vt-le/astnet.git
cd ASTNet/ASTNet

Data preparation

We evaluate ASTNet on:

A dataset is a directory with the following structure:

$ tree data
ped2/avenue
├── training
│   └── frames
│       ├── ${video_1}$
│       │   ├── 000.jpg
│       │   ├── 001.jpg
│       │   └── ...
│       ├── ${video_2}$
│       │   ├── 00.jpg
│       │   └── ...
│       └── ...
├── testing
│   └── frames
│       ├── ${video_1}$
│       │   ├── 000.jpg
│       │   ├── 001.jpg
│       │   └── ...
│       ├── ${video_2}$
│       │   ├── 000.jpg
│       │   └── ...
│       └── ...
└── ped2/avenue.mat

shanghaitech
├── training
│   └── frames
│       ├── ${video_1}$
│       │   ├── 000.jpg
│       │   ├── 001.jpg
│       │   └── ...
│       ├── ${video_2}$
│       │   ├── 00.jpg
│       │   └── ...
│       └── ...
├── testing
│   └── frames
│       ├── ${video_1}$
│       │   ├── 000.jpg
│       │   ├── 001.jpg
│       │   └── ...
│       ├── ${video_2}$
│       │   ├── 000.jpg
│       │   └── ...
│       └── ...
└── test_frame_mask
    ├── 01_0014.npy
    ├── 01_0015.npy
    └── ...

Evaluation

Please first download the pre-trained model

Dataset Pretrained Model
UCSD Ped2 github / drive
CUHK Avenue github / drive
ShanghaiTech github / drive

To evaluate a pretrained ASTNet on a dataset, run:

 python test.py \
    --cfg <path/to/config/file> \
    --model-file </path/to/pre-trained/model>

For example, to evaluate ASTNet on Ped2:

python test.py \
    --cfg config/ped2_wresnet.yaml \
    --model-file pretrained.ped2.pth

Training from scratch

To train ASTNet on a dataset, run:

python train.py \
    --cfg <path/to/config/file>

For example, to train ASTNet on Ped2:

python train.py \
    --cfg config/ped2_wresnet.yaml

Notes:

  • To change other options, see <config/config_file.yaml>.

Citing

If you find our work useful for your research, please consider citing:

@article{le2023attention,
  title={Attention-based Residual Autoencoder for Video Anomaly Detection},
  author={Le, Viet-Tuan and Kim, Yong-Guk},
  journal={Applied Intelligence},
  volume={53},
  number={3},
  pages={3240--3254},
  year={2023},
  publisher={Springer}
}

Contact

For any question, please file an issue or contact:

Viet-Tuan Le: vt-le@outlook.com