(1) Prepare for datasets:
The training and validation data is expected to be in the train folder and val folder respectively, with annotation information saved in csv files:
/path/to/dataset
│── Processed-Frames
│ │── class1
│ ├── img1.jpg
│ ├── class2
│ ├── img2.jpg
│ ├── class3
│ ├── img3.jpg
│
│── train.csv
│── test.csv
(2) Configuration
$ pip install -r requirements.txt$ python -m main.py --config ./configs/UCF_Crime_Image.json --VE vit_b_16 --TRE roberta_base --FD FD --VCE vision_clip --TCE text_clip --save_dir ./output(1) Add your encoder components in model.py
(2) Run the script
$ python -m main.py --config ./configs/TOUR_CONFIG --VE YOUR_VISION_ENCODER --TRE YOUR_TEXT_ENCODER --FD YOUR_DATA_ENCODER --VCE YOUR_VISION_ENCODER2 --TCE YOUR_TEXT_ENCODER2 --save_dir ./output