Dataset
aka Common Objects in Context
COCO
Microsoft's 330k-image dataset with bounding boxes, segmentation and captions.
Definition
COCO is the standard benchmark for object detection, segmentation and captioning. It contains 330k images across 80 object categories with 1.5M instance segmentations. mAP on COCO is the headline number reported for nearly every detection model.
Common use cases
- Object detection
- Segmentation
- Image captioning