On Pursuit of Designing Multi-modal Transformer for Video Grounding

Meng Cao, Long Chen, Mike Zheng Shou, Can Zhang, Yuexian Zou


Abstract
Video grounding aims to localize the temporal segment corresponding to a sentence query from an untrimmed video. Almost all existing video grounding methods fall into two frameworks: 1) Top-down model: It predefines a set of segment candidates and then conducts segment classification and regression. 2) Bottom-up model: It directly predicts frame-wise probabilities of the referential segment boundaries. However, all these methods are not end-to-end, i.e., they always rely on some time-consuming post-processing steps to refine predictions. To this end, we reformulate video grounding as a set prediction task and propose a novel end-to-end multi-modal Transformer model, dubbed as GTR. Specifically, GTR has two encoders for video and language encoding, and a cross-modal decoder for grounding prediction. To facilitate the end-to-end training, we use a Cubic Embedding layer to transform the raw videos into a set of visual tokens. To better fuse these two modalities in the decoder, we design a new Multi-head Cross-Modal Attention. The whole GTR is optimized via a Many-to-One matching loss. Furthermore, we conduct comprehensive studies to investigate different model design choices. Extensive results on three benchmarks have validated the superiority of GTR. All three typical GTR variants achieve record-breaking performance on all datasets and metrics, with several times faster inference speed.
Anthology ID:
2021.emnlp-main.773
Original:
2021.emnlp-main.773v1
Version 2:
2021.emnlp-main.773v2
Volume:
Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2021
Address:
Online and Punta Cana, Dominican Republic
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
9810–9823
Language:
URL:
https://aclanthology.org/2021.emnlp-main.773
DOI:
10.18653/v1/2021.emnlp-main.773
Bibkey:
Cite (ACL):
Meng Cao, Long Chen, Mike Zheng Shou, Can Zhang, and Yuexian Zou. 2021. On Pursuit of Designing Multi-modal Transformer for Video Grounding. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 9810–9823, Online and Punta Cana, Dominican Republic. Association for Computational Linguistics.
Cite (Informal):
On Pursuit of Designing Multi-modal Transformer for Video Grounding (Cao et al., EMNLP 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.emnlp-main.773.pdf
Video:
 https://aclanthology.org/2021.emnlp-main.773.mp4
Data
ActivityNet CaptionsCharades-STA