Poster
Learning Video Representations From Large Language Models
Yue Zhao · Ishan Misra · Philipp Krähenbühl · Rohit Girdhar
West Building Exhibit Halls ABC 235
We introduce LAVILA, a new approach to learning video-language representations by leveraging Large Language Models (LLMs). We repurpose pre-trained LLMs to be conditioned on visual input, and finetune them to create automatic video narrators. Our auto-generated narrations offer a number of advantages, including dense coverage of long videos, better temporal synchronization of the visual information and text, and much higher diversity of text. The video-language embedding learned contrastively with these narrations outperforms the previous state-of-the-art on multiple first-person and third-person video tasks, both in zero-shot and finetuned setups. Most notably, LAVILA obtains an absolute gain of 10.1% on EGTEA classification and 5.9% Epic-Kitchens-100 multi-instance retrieval benchmarks. Furthermore, LAVILA trained with only half the narrations from the Ego4D dataset outperforms models trained on the full set, and shows positive scaling behavior on increasing pre-training data and model size.