sLdZ0306 / CCR

Counterfactual Cross-modality Reasoning for Weakly Supervised Video Moment Localization

CCR

Implementaion of the paper "Counterfactual Cross-modality Reasoning for Weakly Supervised Video Moment Localization" (ACM MM 2023).

Data preparation

Features for Charades-STA (by LGI) and for ActivityNet Captions. Download these features and put them into the corresponding directories in "\data". Refering to CPL for more details.

Training and inference

Train

# Charades-STA
python train.py --config-path config/charades/main.json --log_dir LOG_DIR --tag TAG --vote
# ActivityNet Captions
python train.py --config-path config/activitynet/main.json --log_dir LOG_DIR --tag TAG

Inference

python train.py --config-path CONFIG_FILE --resume CHECKPOINT_FILE --eval