v1.0

The following files contains the MOVAD pretrained model on the Detection of Traffic Anomaly (DoTA) dataset.

To use pretrained models and/or test fast evaluation of the models themself (loading the results.pkl file), unpack files as show here:

output/
├── v4_1
│   ├── checkpoints
│   │   └── model-640.pt
│   └── eval
│       └── results-640.pkl
└── v4_2
    ├── checkpoints
    │   └── model-690.pt
    └── eval
        └── results-690.pkl

As written inside the README, we recap how you can run the training/eval:

Train

python src/main.py --config cfgs/v4_2.yml --output output/v4_2/ --phase train --epochs 1000 --epoch -1

Setting --epoch -1, the code will load latest available checkpoint (model-690.pt in this case) automatically.

Eval

python src/main.py --config cfgs/v4_2.yml --output output/v4_2/ --phase test --epoch 690

Play: generate video

python src/main.py --config cfgs/v4_2.yml --output output/v4_2/ --phase play --epoch 690