80 object categories and a reduced set of images from the COCO 2017 collection are provided for object detection and segmentation tasks. This dataset serves as a smaller-scale version of the standard benchmark to facilitate faster iteration in computer vision workflows.
Use Cases
- Train object detection models using the bbox and category_id columns
- Develop instance segmentation algorithms using the segmentation polygon data
- Test data loading pipelines designed for the COCO JSON annotation format
Strengths
- Includes 80 object categories such as 'person', 'car', and 'dog'
- Provides bounding box data in the [x, y, width, height] format
- Contains segmentation masks for each annotated object instance