Crafting a Toolchain for Image Restoration by Deep Reinforcement Learning

Citations Over TimeTop 10% of 2018 papers

Abstract

We investigate a novel approach for image restoration by reinforcement learning. Unlike existing studies that mostly train a single large network for a specialized task, we prepare a toolbox consisting of small-scale convolutional networks of different complexities and specialized in different tasks. Our method, RL-Restore, then learns a policy to select appropriate tools from the toolbox to progressively restore the quality of a corrupted image. We formulate a stepwise reward function proportional to how well the image is restored at each step to learn the action policy. We also devise a joint learning scheme to train the agent and tools for better performance in handling uncertainty. In comparison to conventional human-designed networks, RL-Restore is capable of restoring images corrupted with complex and unknown distortions in a more parameter-efficient manner using the dynamically formed toolchain 1 .

Related Papers

• → The LSTS toolchain for networked vehicle systems(2013)138 cited
• → Software Numerical Instability Detection and Diagnosis by Combining Stochastic and Infinite-Precision Testing(2016)11 cited
• → Stream machine learning on vehicle data(2018)2 cited
• → SKilLed communication for toolchains(2015)
• → Finite Element Method Based Toolchain for Simulation of Proximity Estimation Using Electronic Skin(2023)