3D-FRONT-FUTURE / 3D-FUTURE-ToolBox

FUTURE3D Toolbox: Rendering, Projection, and Re-Projection
https://tianchi.aliyun.com/specials/promotion/ijcai-alibaba-3d-future-workshop?spm=5176.12281976.0.0.1487178eSSlGKi
125 stars 18 forks source link

object semantic annotations for scene image #4

Closed WenM1222 closed 3 years ago

WenM1222 commented 3 years ago

Hi, In 3D-FUTURE-scene\GT\train_set.json, is the list in node annotations\segmentation\counts for getting the segmented object contour? Could you please to explain how to use it? It looks not like coordinate, as some value is over 1200 (scene image size is 1200 * 1200) and the number of point is a singular. If not, what's the usage of the list? I couldn't find any explanation. Thank you !

WenM1222 commented 3 years ago

The annotation format follows the coco dataset.