Dynamic Scene Understanding from Vision-Language Representations
This work addresses the problem of dynamic scene understanding for computer vision researchers, offering an incremental improvement by applying existing representations to new tasks.
The paper tackles the challenge of automatically parsing complex, dynamic scenes by leveraging frozen vision-language representations to unify tasks like Situation Recognition and Human-Human/Human-Object Interactions, achieving state-of-the-art results with minimal trainable parameters.
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.