Home

Awesome

CSMGAN

Code for ACM MM2020 paper

Jointly Cross- and Self-Modal Graph Attention Network for Query-Based Moment Localization <br /> [Paper] <br />

Main Results

Activity Caption
R@1, IoU=0.3R@1, IoU=0.5R@1, IoU=0.7R@5, IoU=0.3R@5, IoU=0.5R@5, IoU=0.7
68.5249.1129.1587.6877.4359.63
TACoS
R@1, IoU=0.1R@1, IoU=0.3R@1, IoU=0.5R@5, IoU=0.1R@5, IoU=0.3R@5, IoU=0.5
42.7433.9027.0968.9753.9841.22
Charades-STA
R@1, IoU=0.5R@1, IoU=0.7R@5, IoU=0.5R@5, IoU=0.7
60.0437.3489.0161.85
DiDeMo
R@1, IoU=0.5R@1, IoU=0.7R@5, IoU=0.5R@5, IoU=0.7
29.4419.1670.7741.61

Prerequisites

Preparation

Evaluation

$ python main.py --word2vec-path /yourpath/glove_model.bin --dataset ActivityNet --feature-path /yourpath/ActivityCaptions/ActivityC3D --train-data data/activity/train_data_gcn.json --val-data data/activity/val_data_gcn.json --test-data data/activity/test_data_gcn.json --max-num-epochs 20 --dropout 0.2 --warmup-updates 300 --warmup-init-lr 1e-06 --lr 8e-4 --num-heads 4 --num-gcn-layers 2 --num-attn-layers 2 --weight-decay 1e-7 --evaluate --model-load-path ./models_activity/model_6852
$ python main.py --word2vec-path /yourpath/glove_model.bin --dataset TACOS --feature-path /yourpath/TACOS/TACOS --train-data data/tacos/TACOS_train_gcn.json --val-data data/tacos/TACOS_val_gcn.json --test-data data/tacos/TACOS_test_gcn.json --max-num-epochs 40 --dropout 0.2 --warmup-updates 300 --warmup-init-lr 1e-07 --lr 4e-4 --num-heads 4 --num-gcn-layers 2 --num-attn-layers 2 --weight-decay 1e-8 --evaluate --model-saved-path models_tacos --batch-size 64 --model-load-path ./models_tacos/model_4274

Citation

If you use this code please cite:

@inproceedings{liu2020jointly,
    title={Jointly Cross- and Self-Modal Graph Attention Network for Query-Based Moment Localization},
    author={Liu, Daizong and Qu, Xiaoye and Liu, Xiaoyang and Dong, Jianfeng and Zhou, Pan and Xu, Zichuan},
    booktitle={Proceedings of the 28th ACM International Conference on Multimedia (MM’20)},
    year={2020}
}