Removing Rain Streaks via Task Transfer Learning
Yinglong Wang, Chao Ma, Jianzhuang Liu

TL;DR
This paper introduces a task transfer learning approach for image deraining that leverages related tasks and real data to improve generalization to real-world rainy scenes, outperforming existing methods.
Contribution
Proposes a novel task transfer learning framework with knowledge distillation to enhance deraining generalization and reduce model size.
Findings
Outperforms state-of-the-art supervised deraining methods on synthetic data.
Shows superior generalization to real-world rainy scenes.
Effectively reduces model size via knowledge distillation.
Abstract
Due to the difficulty in collecting paired real-world training data, image deraining is currently dominated by supervised learning with synthesized data generated by e.g., Photoshop rendering. However, the generalization to real rainy scenes is usually limited due to the gap between synthetic and real-world data. In this paper, we first statistically explore why the supervised deraining models cannot generalize well to real rainy cases, and find the substantial difference of synthetic and real rainy data. Inspired by our studies, we propose to remove rain by learning favorable deraining representations from other connected tasks. In connected tasks, the label for real data can be easily obtained. Hence, our core idea is to learn representations from real data through task transfer to improve deraining generalization. We thus term our learning strategy as \textit{task transfer learning}.…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsImage Enhancement Techniques · Video Surveillance and Tracking Methods · Advanced Image Fusion Techniques
