Learning Across Tasks and Domains
This work addresses the challenge of practical knowledge transfer across tasks and domains in computer vision, providing incremental improvements over existing domain adaptation techniques.
The paper tackles the problem of transferring learned concepts across different tasks and domains by introducing a novel adaptation framework that operates across both, achieving performance gains in monocular depth estimation and semantic segmentation across four domains.
Recent works have proven that many relevant visual tasks are closely related one to another. Yet, this connection is seldom deployed in practice due to the lack of practical methodologies to transfer learned concepts across different training processes. In this work, we introduce a novel adaptation framework that can operate across both task and domains. Our framework learns to transfer knowledge across tasks in a fully supervised domain (e.g., synthetic data) and use this knowledge on a different domain where we have only partial supervision (e.g., real data). Our proposal is complementary to existing domain adaptation techniques and extends them to cross tasks scenarios providing additional performance gains. We prove the effectiveness of our framework across two challenging tasks (i.e., monocular depth estimation and semantic segmentation) and four different domains (Synthia, Carla, Kitti, and Cityscapes).