cvpr2018最佳paper，其目标在于利用视觉任务之间的关联性减少总体数据使用量。
感兴趣的问题：
-
哪26个任务
- surface normals
- Eucl.distance
- object class.
- scene class.
- jigsaw puzzle
- colorization
- 2D segm.
- 2.5D segm.
- semantic segm.
- valishing points
- 2D edges
- 3D edges
- 2D keypoints
- 3D keypoints
- 3D curvature
- image reshading
- in-painting
- denoising
- autoencoding
- cam. pos(non fixed)
- cam. pos(fixed)
- triplet cam pose
- room layout
- point matching
- random projection
-
Egomotion(帧间运动)
tasks.png
-
与semantic segmentation相关的任务有哪些
- object class.
-
scene class.
task tree.png
-
网络结构：类ResNet50
网友评论