Dynamic Scene Understanding from Vision-Language Representations

TL;DR

This paper introduces a framework that leverages modern vision-language representations to understand complex, dynamic scenes, achieving state-of-the-art results with minimal task-specific engineering and trainable parameters.

Contribution

It demonstrates how frozen vision-language models can be used for dynamic scene understanding tasks by framing them as structured text prediction or representation concatenation, reducing the need for extensive training.

Findings

State-of-the-art performance on dynamic scene understanding tasks

Modern V&L representations encode dynamic scene semantics effectively

Minimal trainable parameters needed for high performance

Abstract

Images depicting complex, dynamic scenes are challenging to parse automatically, requiring both high-level comprehension of the overall situation and fine-grained identification of participating entities and their interactions. Current approaches use distinct methods tailored to sub-tasks such as Situation Recognition and detection of Human-Human and Human-Object Interactions. However, recent advances in image understanding have often leveraged web-scale vision-language (V&L) representations to obviate task-specific engineering. In this work, we propose a framework for dynamic scene understanding tasks by leveraging knowledge from modern, frozen V&L representations. By framing these tasks in a generic manner - as predicting and parsing structured text, or by directly concatenating representations to the input of existing models - we achieve state-of-the-art results while using a minimal number of trainable parameters relative to existing approaches. Moreover, our analysis of dynamic knowledge of these representations shows that recent, more powerful representations effectively encode dynamic scene semantics, making this approach newly possible.