PRISM：零样本免训练的程序性视频摘要框架，仅选 5% 帧保留 84% 语义

View PDF HTML (experimental)

Abstract:Video summarization transforms long videos into concise representations that are easier to document and analyze, especially in high-stakes domains such as surgical training. However, long-form videos often require dense frame processing or supervised training pipelines, which can be computationally expensive and may still miss important procedural content. We present PRISM: Procedural Representation via Integrated Semantic and Multimodal Analysis, a zero-shot and training-free framework for frame-efficient procedural video summarization. PRISM selects fewer than 5% of video frames while retaining over 84% semantic content and improving over baselines by up to 7.5%. Rather than exhaustively processing frames, PRISM uses lightweight visual filtering and dynamically generated procedural labels as semantic anchors to select meaningful frame-label groups. This selective inference design preserves key actions, transitions, and contextual details while reducing the number of visual inputs passed to downstream vision-language captioning stages. We evaluate PRISM on YouCook2 and ActivityNet Captions, with additional studies on keyframe selection benchmarks and surgical video datasets. Across procedural and domain-specific video tasks, PRISM achieves strong semantic alignment and accuracy, suggesting that efficient multimodal video understanding can be achieved by grounding generation in dynamically generated semantic anchors.

Comments:	18 pages, 6 figures
Subjects:	Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
Cite as:	arXiv:2601.12243 [cs.CV]
	(or arXiv:2601.12243v3 [cs.CV] for this version)
	https://doi.org/10.48550/arXiv.2601.12243 arXiv-issued DOI via DataCite

Submission history

From: Shreya Rajpal [view email]
[v1] Sun, 18 Jan 2026 03:41:48 UTC (2,548 KB)
[v2] Fri, 30 Jan 2026 00:36:33 UTC (2,548 KB)
[v3] Sun, 11 Oct 2026 02:06:57 UTC (2,502 KB)