NAVERO: Unlocking Fine-Grained Semantics for Video-Language Compositionality
This addresses the problem of compositional understanding in video-language models for AI and computer vision applications, representing an incremental advance with a novel method for a known bottleneck.
The paper tackled the challenge of video-language models understanding fine-grained compositions like objects, attributes, and actions that change over time, by proposing NAVERO, a training method using negative text augmentation, which achieved significant improvements in compositional understanding and maintained strong retrieval performance.
We study the capability of Video-Language (VidL) models in understanding compositions between objects, attributes, actions and their relations. Composition understanding becomes particularly challenging for video data since the compositional relations rapidly change over time in videos. We first build a benchmark named AARO to evaluate composition understanding related to actions on top of spatial concepts. The benchmark is constructed by generating negative texts with incorrect action descriptions for a given video and the model is expected to pair a positive text with its corresponding video. Furthermore, we propose a training method called NAVERO which utilizes video-text data augmented with negative texts to enhance composition understanding. We also develop a negative-augmented visual-language matching loss which is used explicitly to benefit from the generated negative text. We compare NAVERO with other state-of-the-art methods in terms of compositional understanding as well as video-text retrieval performance. NAVERO achieves significant improvement over other methods for both video-language and image-language composition understanding, while maintaining strong performance on traditional text-video retrieval tasks.