Search papers, labs, and topics across Lattice.
This paper introduces POS-ISP, a sequence-level reinforcement learning framework for optimizing image signal processing (ISP) pipelines. POS-ISP predicts the entire module sequence and parameters in a single forward pass, enabling end-to-end optimization based on a terminal task reward. Experiments demonstrate that POS-ISP achieves improved task performance and reduced computational cost compared to existing NAS and step-wise RL approaches.
End-to-end reinforcement learning can discover better image processing pipelines than step-wise methods, while also being more computationally efficient.
Recent work has explored optimizing image signal processing (ISP) pipelines for various tasks by composing predefined modules and adapting them to task-specific objectives. However, jointly optimizing module sequences and parameters remains challenging. Existing approaches rely on neural architecture search (NAS) or step-wise reinforcement learning (RL), but NAS suffers from a training-inference mismatch, while step-wise RL leads to unstable training and high computational overhead due to stage-wise decision-making. We propose POS-ISP, a sequence-level RL framework that formulates modular ISP optimization as a global sequence prediction problem. Our method predicts the entire module sequence and its parameters in a single forward pass and optimizes the pipeline using a terminal task reward, eliminating the need for intermediate supervision and redundant executions. Experiments across multiple downstream tasks show that POS-ISP improves task performance while reducing computational cost, highlighting sequence-level optimization as a stable and efficient paradigm for task-aware ISP. The project page is available at https://w1jyun.github.io/POS-ISP
