Event
Taskonomy: Modeling the Transfer Dependency Structure of Visual Tasks
1 reports · 1 sources
Amir Zamir et al. proposed Taskonomy, which uses a fully computational approach to model 26 2D, 2.5D, 3D, and semantic visual tasks in a latent space, generating a task taxonomy graph by discovering first-order and higher-order transfer learning dependencies among tasks. The authors claim that the total amount of labeled data required to solve a set of 10 tasks can be reduced by about two-thirds compared with training each independently, while performance remains essentially on par; they also provide a set of tools for computing and probing this structure, including a solver for users to formulate efficient supervision strategies. The paper received the CVPR 2018 Best Paper Award.
Report timeline
- Taskonomy: Disentangling Task Transfer Learning
Taskonomy proposes a fully computational method that models 26 2D, 2.5D, 3D, and semantic visual tasks in a latent space, mines first-order and higher-order transfer learning dependencies, and produces a taxonomic map for task transfer learning.
arXiv cs.RO ·