The goal of this joint challenge of COCO and Places is to bring the question of object recognition in the context of the scene understanding.
COCO is an image dataset designed to spur object detection research with a focus on full scene understanding. The annotations include: presence of objects belonging to 80 common categories, pixel-level segmentation of each object, keypoint annotations for person instances, five image captions per image. The specific tracks in the COCO Challenge 2017 are (1) object detection with bounding boxes, (2) object detection with segmentation masks, and (3) joint detection and person keypoint estimation. See the following COCO challenge page for the detailed information:
Places Challenge will host three tracks meant to complement the COCO Challenge. The data the Places Challenge 2017 are from the pixel-wise annotated image dataset ADE20K, in which there are 20K images for training, 2K images for validation, and 3K images for testing. The three specific tracks in the Places Challenge 2017 are: (1) scene parsing, (2) instance segmentation, and (3) semantic boundary detection. See the following Places Challenge page for the detailed information:
- June 25, 2017: Development kits and data are available.
- Sep.15, 2017: Submission deadline for Places Challenge (Check Places Challenge page for the latest news)
- Sep.30, 2017: Submission deadline for COCO Challenge (Check COCO challenge page for the latest news)
- Sept.26, 2017: Places Challenge results released.
- Oct.22, 2017: COCO Challenge winners notified. Results officially released at the COCO workshop.
- Oct.29, 2017: Joint COCO and Places Challenge Workshop (full-day)
To be announced
- Raquel Urtasun, University of Toronto
- Vladlen Koltun, Intel Labs