ActivityNet
video
+
+
+
+
About
ActivityNet is a large-scale video understanding benchmark featuring 203 complex human activity classes with 849 hours of untrimmed YouTube videos. Designed for human activity recognition research, it provides 19,994 videos with temporal annotations for three evaluation tasks: untrimmed video classification, trimmed activity classification, and activity detection. The dataset covers diverse real-world activities from daily living, making it the largest temporal activity detection benchmark.
+
+
+
+
Evaluation Stats
Total Models1
Organizations1
Verified Results0
Self-Reported1
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers
Score Distribution
1 models
Top Score
61.9%
Average Score
61.9%
High Performers (80%+)
0Top Organizations
#1OpenAI
1 model
61.9%
+
+
+
+
Leaderboard
1 models ranked by performance on ActivityNet
