Learning Good Features to Transfer Across Tasks and Domains

TL;DR

This paper introduces a method to transfer learned features across different tasks and domains in computer vision by learning a generalizable mapping between task-specific features, improving performance in synthetic-to-real adaptation scenarios.

Contribution

It proposes a neural network-based feature mapping approach that generalizes across unseen domains and tasks, with strategies to enhance learning and generalization capabilities.

Findings

Effective transfer of knowledge between monocular depth estimation and semantic segmentation.

Significant improvements in synthetic-to-real domain adaptation.

Enhanced generalization through constrained feature spaces.

Abstract

Availability of labelled data is the major obstacle to the deployment of deep learning algorithms for computer vision tasks in new domains. The fact that many frameworks adopted to solve different tasks share the same architecture suggests that there should be a way of reusing the knowledge learned in a specific setting to solve novel tasks with limited or no additional supervision. In this work, we first show that such knowledge can be shared across tasks by learning a mapping between task-specific deep features in a given domain. Then, we show that this mapping function, implemented by a neural network, is able to generalize to novel unseen domains. Besides, we propose a set of strategies to constrain the learned feature spaces, to ease learning and increase the generalization capability of the mapping network, thereby considerably improving the final performance of our framework. Our proposal obtains compelling results in challenging synthetic-to-real adaptation scenarios by transferring knowledge between monocular depth estimation and semantic segmentation tasks.