Multi-modal integrated proposal generation network for weakly supervised video moment retrieval

Dikai Fang, Huahu Xu, Wei Wei, Mohsen Guizani, Honghao Gao

Abstract: Highlights•Propose MIPGN: A novel approach for improved video moment retrieval.•Frame clustering acquires scene context, guiding adaptive proposal generation.•Leverage pretrained models for multi-modal tag extraction, enriching feature representation.•Contrastive learning-based multi-task objective boosts model training efficiency.•Efficacy demonstrated through testing on Charades-STA and ActivityNet datasets.