VideoPrism: A Foundational Visual Encoder for Video Understanding

Autor:	Zhao, Long, Gundavarapu, Nitesh B., Yuan, Liangzhe, Zhou, Hao, Yan, Shen, Sun, Jennifer J., Friedman, Luke, Qian, Rui, Weyand, Tobias, Zhao, Yue, Hornung, Rachel, Schroff, Florian, Yang, Ming-Hsuan, Ross, David A., Wang, Huisheng, Adam, Hartwig, Sirotenko, Mikhail, Liu, Ting, Gong, Boqing
Rok vydání:	2024
Předmět:	Computer Science - Computer Vision and Pattern Recognition Computer Science - Artificial Intelligence
Druh dokumentu:	Working Paper
Popis:	We introduce VideoPrism, a general-purpose video encoder that tackles diverse video understanding tasks with a single frozen model. We pretrain VideoPrism on a heterogeneous corpus containing 36M high-quality video-caption pairs and 582M video clips with noisy parallel text (e.g., ASR transcripts). The pretraining approach improves upon masked autoencoding by global-local distillation of semantic video embeddings and a token shuffling scheme, enabling VideoPrism to focus primarily on the video modality while leveraging the invaluable text associated with videos. We extensively test VideoPrism on four broad groups of video understanding tasks, from web video question answering to CV for science, achieving state-of-the-art performance on 31 out of 33 video understanding benchmarks. Comment: Accepted to ICML 2024. v2: added retrieval results on MSRVTT (1K-A), more data analyses, and ablation studies
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/2402.13217 Zobrazit plný text záznamu View this record from Arxiv