SOVABench: A Vehicle Surveillance Action Retrieval Benchmark for Multimodal Large Language Models

TL;DR

SOVABench is a new surveillance-focused video retrieval benchmark testing vehicle action discrimination and temporal understanding, revealing current models' challenges and demonstrating a training-free MLLM-based approach for interpretable embeddings.

Contribution

The paper introduces SOVABench, a novel real-world surveillance benchmark for vehicle action retrieval, and proposes a training-free MLLM-based framework for interpretable video and image embeddings.

Findings

State-of-the-art models struggle with cross-action discrimination.

MLLM-based embeddings outperform contrastive vision-language models on several benchmarks.

SOVABench provides a challenging testbed for surveillance action retrieval.

Abstract

Automatic identification of events and recurrent behavior analysis are critical for video surveillance. However, most existing content-based video retrieval benchmarks focus on scene-level similarity and do not evaluate the action discrimination required in surveillance. To address this gap, we introduce SOVABench (Surveillance Opposite Vehicle Actions Benchmark), a real-world retrieval benchmark built from surveillance footage and centered on vehicle-related actions. SOVABench defines two evaluation protocols (inter-pair and intra-pair) to assess cross-action discrimination and temporal direction understanding. Although action distinctions are generally intuitive for human observers, our experiments show that they remain challenging for state-of-the-art vision and multimodal models. Leveraging the visual reasoning and instruction-following capabilities of Multimodal Large Language Models (MLLMs), we present a training-free framework for producing interpretable embeddings from MLLM-generated descriptions for both images and videos. The framework achieves strong performance on SOVABench as well as on several spatial and counting benchmarks where contrastive Vision-Language Models often fail. The code, annotations, and instructions to construct the benchmark are publicly available.