Poster
: Towards Analyzing Fast, Frequent, and Fine-grained Events from Videos
Zhaoyu Liu · Kan Jiang · Murong Ma · Zhe Hou · Yun Lin · Jin Song Dong
[
Abstract
]
Thu 24 Apr midnight PDT
— 2:30 a.m. PDT
Abstract:
Analyzing Fast, Frequent, and Fine-grained () events presents a significant challenge in video analytics and multi-modal LLMs. Current methods struggle to identify events that satisfy all the criteria with high accuracy due to challenges such as motion blur and subtle visual discrepancies. To advance research in video understanding, we introduce , a benchmark that consists of video datasets for precise event detection. Datasets in are characterized by their extensive scale and comprehensive detail, usually encompassing over 1,000 event types with precise timestamps and supporting multi-level granularity. Currently, contains several sports datasets, and this framework may be extended to other applications as well. We evaluated popular temporal action understanding methods on , revealing substantial challenges for existing techniques. Additionally, we propose a new method, , for event detections, achieving superior performance. The dataset, model, and benchmark code are available at https://github.com/F3Set/F3Set.
Live content is unavailable. Log in and register to view live content