Video · Image / Education
Fight Detection Video Dataset - 3,010 Videos for Public Safety and Computer Vision
Advertised by providerLicensing rights not independently verifiedPaid licence
Overview
Fight Detection Video Dataset - 3,010 Videos for Public Safety and Computer Vision is a video and image dataset. The provider states a volume of 3,010 videos. Typical applications include alignment & RLHF and object detection. Relevant sectors: Education, AI & Machine Learning and Public Sector & Legal. It is offered as a paid dataset; pricing details are provided on request.
Primary applications
What the listing offers
- Stated volume of 3,010 videos.
Overview The Fight Detection Video Dataset is a commercial video dataset produced by the provider. It contains 3,010 videos of simulated physical altercations involving 2 to 4 participants across 1v1, 2v1, 2v2 and mass brawl scenarios. The videos were captured in controlled environments from static and moving camera perspectives at 1920×1080 resolution and 30 FPS, across approximately 50–60 locations. The dataset is designed for violence detection, fighting detection, action recognition and public safety systems. Dataset at a Glance 1. Total number of files: 3,010 videos 2. Data type: video 3. Video formats: MP4, MOV 4. Resolution: 1920×1080 (Full HD) 5. Frame rate: 30 FPS 6. Number of cameras: 2 7. Camera perspectives: static and moving 8. Scenarios: 1v1, 2v1, 2v2, mass brawl 9. Participants count: 2 to 4 10. Number of locations: ~50–60 11. Clothing change: every 10 videos 12. Tasks: public safety, computer vision Fight Scenarios The dataset covers four types of fighting scenes, from one-on-one conflicts to group activity: - 1v1: physical violence between two participants - 2v1: two participants against one - 2v2: group fight between two pairs - Mass brawl: group violence with multiple participants This range of violent scenarios helps detection systems recognize violent interactions at different scales. It also supports research into crowd behavior and violence in crowded scenes. Number of People Distribution Videos are almost evenly balanced by participant count: - 2 people: 99
Related datasets
More in Video & Physical AI →Video · ImageAdvertised
Abandoned Object Detection Video Dataset - 1,062 Videos for Public Safety and Computer Vision
Overview The Abandoned Object Detection Video Dataset (referred to on the site as the ABODA dataset) is a commercial video dataset produced by the provider.
- Language
- —
- Volume
- 1,062 videos
- Domain
- Education +2
- Use case
- Alignment & RLHF
USD 30,000
More video & physical ai datasets
All 302 video & physical ai datasets →- First-Person Video Egocentric Data : Agriculture, Farming & Landscaping Tasks | 157°/84° FOV Stereo Camera + Motion Sensors | Embodied AI Training
- First-Person Video: Plastics & Light-Assembly Tasks | Dual Stereo Camera, Hardware-Synced IMU | Embodied AI Training for Deep Learning (DL) Data
- First-Person Woodworking & Furniture-Assembly Video | Hand-Keypoint & Camera-Head Trajectory Data | Embodied Data & AI, Robot Foundation Models
- French documentaries spanning history, science, nature, culture
- French media content spanning history, science, culture, sports
- Full Library | 50,000 4K Video Clips | 1,750+ Hours | 360-degree POV Driving Footage
- GELLO Bimanual Teleop — 5 Golden Dataset
- German animation video across diverse genres