YouTube-BoundingBoxes: A Large High-Precision Human-Annotated Data Set for Object Detection in Video.
Esteban RealJonathon ShlensStefano MazzocchiXin PanVincent VanhouckePublished in: CVPR (2017)
Keyphrases
- high precision
- object detection
- data sets
- high recall
- video sharing
- multimedia
- achieve high precision
- high reliability
- video sequences
- user generated
- video search
- youtube videos
- video streams
- video dataset
- high accuracy
- video data
- face detection
- video frames
- video content
- video surveillance
- scene understanding
- video analysis
- pedestrian detection
- motion capture data
- video clips
- key frames
- object recognition
- static images
- social media
- action detection
- real time
- human activities
- action recognition
- human subjects
- video retrieval
- visual saliency
- face recognition
- multimedia data
- computer vision
- online video
- video material
- real world
- multi class
- spatio temporal