English

Generative Modeling for Multi-task Visual Learning

Computer Vision and Pattern Recognition 2021-06-28 v1

Abstract

Generative modeling has recently shown great promise in computer vision, but it has mostly focused on synthesizing visually realistic images. In this paper, motivated by multi-task learning of shareable feature representations, we consider a novel problem of learning a shared generative model that is useful across various visual perception tasks. Correspondingly, we propose a general multi-task oriented generative modeling (MGM) framework, by coupling a discriminative multi-task network with a generative network. While it is challenging to synthesize both RGB images and pixel-level annotations in multi-task scenarios, our framework enables us to use synthesized images paired with only weak annotations (i.e., image-level scene labels) to facilitate multiple visual tasks. Experimental evaluation on challenging multi-task benchmarks, including NYUv2 and Taskonomy, demonstrates that our MGM framework improves the performance of all the tasks by large margins, consistently outperforming state-of-the-art multi-task approaches.

Keywords

Cite

@article{arxiv.2106.13409,
  title  = {Generative Modeling for Multi-task Visual Learning},
  author = {Zhipeng Bao and Martial Hebert and Yu-Xiong Wang},
  journal= {arXiv preprint arXiv:2106.13409},
  year   = {2021}
}