keyword

long-range video understanding

Long-range video understanding is a branch of computer vision and artificial intelligence dedicated to analyzing, interpreting, and reasoning about extended video content that spans minutes, hours, or entire narratives. Unlike traditional video recognition methods that process short, few-second clips to identify isolated actions, long-range video understanding models track complex temporal dependencies, evolving storylines, and contextual relationships across multiple shots, scenes, and events over prolonged durations. Developing these systems requires computational architectures designed to overcome memory and processing bottlenecks associated with long frame sequences, enabling tasks such as scene boundary detection, video summarization, temporal event localization, and long-form narrative comprehension.

1 item

Efficient Movie Scene Detection using State-Space Transformers

Md Mohaiminul Islam, Mahmudul Hasan, Kishan Shamsundar Athrey, Tony Braskich, Gedas Bertasius

OrganizationsComcastUniversity of North Carolina at Chapel Hill

Why you should read this

Proposes TranS4mer, a hybrid architecture combining structured state-space sequence modeling with self-attention to detect movie scene boundaries across long video sequences while cutting GPU memory usage by three times compared to standard transformers.

The ability to distinguish between different movie scenes is critical for understanding the storyline of a movie. However, accurately detecting movie scenes is often challenging as it requires the ability to reason over very long movie segments. This contrasts with most existing video recognition models, which are typically designed for short-range video analysis. This work proposes a State-Space Transformer model that can efficiently capture dependencies in long movie videos for accurate movie scene detection. Our model, called TranS4mer, is built using a novel S4A building block, combining the strengths of structured state-space sequence (S4) and self-attention (A) layers. Given a sequence of frames divided into movie shots (uninterrupted periods where the camera position does not change), the S4A block first applies self-attention to capture short-range intra-shot dependencies. Afterward, the state-space operation in the S4A block aggregates long-range inter-shot cues. The final TranS4mer model, which can be trained end-to-end, is obtained by stacking the S4A blocks one after the other multiple times. Our proposed TranS4mer outperforms all prior methods in three movie scene detection datasets, including MovieNet, BBC, and OVSD, while being 2× faster and requiring 3× less GPU memory than standard Transformer models. We will release our code and models.

Source

https://arxiv.org/pdf/2212.14427

Added

2026-09-26