Senior Research Scientist — Low-level & Physics-based Vision
Soyoung Choi
Current focus: End-to-end differentiable ISP that improves nighttime pedestrian recall by ≥3 mAP over fixed-pipeline ISP.
Background
PhD Princeton CS 2022, advised by Felix Heide at the Computational Imaging Lab; thesis on differentiable ISP pipelines co-optimized with downstream perception. Research Scientist at Sony AI Imaging (2022–2025), co-author of the open-source `diff-isp` and night-driving raw datasets used by several AD teams.
Education
- PhD · Computer Science — Princeton University (2022), advisor: Felix Heide
- BS · Electrical and Computer Engineering — Seoul National University (2016)
Selected publications
- Choi & Heide, 'Differentiable ISPs for Joint Perception Optimization', CVPR 2022 (oral)
- Choi et al., 'Physics-Based HDR Denoising for Night Driving', ICCV 2023
- Choi et al., 'Learned Demosaicing for Multi-Bayer Automotive Sensors', TPAMI 2024
Reading list
- Chen et al., 'Learning to See in the Dark (SID)', CVPR 2018
- Zamir et al., 'Restormer: Efficient Transformer for High-Resolution Image Restoration', CVPR 2022
- Brooks et al., 'Unprocessing Images for Learned Raw Denoising', CVPR 2019
- Tseng et al., 'Hyperparameter Optimization in Black-box Image Processing using Differentiable Proxies', SIGGRAPH 2019
- Mosleh et al., 'Hardware-in-the-Loop End-to-End Optimization of Camera Image Processing Pipelines', CVPR 2020