Multi-task Learning for Computer Vision
GPTKB entity
Statements (46)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
research
|
| gptkbp:application |
pose estimation
image classification object detection semantic segmentation depth estimation |
| gptkbp:approach |
multi-modal learning
cross-stitch networks hard parameter sharing soft parameter sharing task-specific layers |
| gptkbp:challenge |
balancing losses
negative transfer task interference |
| gptkbp:conference |
gptkb:CVPR
gptkb:ECCV gptkb:ICCV gptkb:NeurIPS |
| gptkbp:database |
gptkb:COCO
gptkb:Cityscapes gptkb:PASCAL_VOC NYU Depth V2 |
| gptkbp:field |
computer vision
|
| gptkbp:goal |
efficient learning
improve generalization reduce overfitting |
| https://www.w3.org/2000/01/rdf-schema#label |
Multi-task Learning for Computer Vision
|
| gptkbp:relatedTo |
gptkb:Mask_R-CNN
self-supervised learning transfer learning representation learning multi-task learning domain adaptation meta-learning Taskonomy Cross-stitch Networks MTAN Sluice Networks UberNet |
| gptkbp:surveyedBy |
Ruder 2017
Zamir et al. 2018 |
| gptkbp:uses |
convolutional neural networks
deep learning shared representations |
| gptkbp:bfsParent |
gptkb:Dr._Alex_Kendall
|
| gptkbp:bfsLayer |
8
|