Benchmark RadarEvent ActivityNet
Event ActivityNet is a large-scale simulated-event benchmark for untrimmed action understanding, derived from ActivityNet videos. It includes 3,263 videos, 200 action classes, event-voxel representations, temporal annotations, and supports recognition, event-language alignment, and online temporal localization.
- Released
- 2026-08-03
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Existing datasets lack long-horizon event-based understanding. Event ActivityNet provides a scalable benchmark for long-horizon event modeling with multiple tasks, enabling evaluation of models on untrimmed action understanding.
Motivation
Long-horizon event-based action understanding remains underexplored because existing datasets largely comprise short, trimmed clips, while collecting native event streams with dense temporal annotations is costly.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.