# Learned Camera Gain and Exposure Control for Improved Visual Feature Detection and Matching

arXiv: https://arxiv.org/abs/2102.04341 · HTML: https://mlsanity.com/paper/2102.04341/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- Builds on 10 of its 25 references that are in the map.
- First posted on arXiv 8 Feb 2021.

## Builds on (most cited first, 10 of 10)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift](https://mlsanity.com/paper/1502.03167/) (2015)
- [ORB-SLAM: a Versatile and Accurate Monocular SLAM System](https://mlsanity.com/paper/1502.00956/) (2015)
- [LIFT: Learned Invariant Feature Transform](https://mlsanity.com/paper/1603.09114/) (2016)
- [Adversarial Training for Adverse Conditions: Robust Metric Localisation using Appearance Transfer](https://mlsanity.com/paper/1803.03341/) (2018)
- [Camera Exposure Control for Robust Robot Vision with Noise-Aware Image Quality Assessment](https://mlsanity.com/paper/1907.12646/) (2019)
- [Learning Matchable Image Transformations for Long-term Metric Visual Localization](https://mlsanity.com/paper/1904.01080/) (2019)
- [Learning-based Image Enhancement for Visual Odometry in Challenging HDR Environments](https://mlsanity.com/paper/1707.01274/) (2017)
- [How to Train a CAT: Learning Canonical Appearance Transformations for Direct Visual Localization Under Illumination Change](https://mlsanity.com/paper/1709.03009/) (2017)

## Built on by (most cited first, 13 of 13)

- [Rawgment: Noise-Accounted RAW Augmentation Enables Recognition in a Wide Variety of Environments](https://mlsanity.com/paper/2210.16046/) (2022)
- [Learning to Control Camera Exposure via Reinforcement Learning](https://mlsanity.com/paper/2404.01636/) (2024)
- [PPISP: Physically-Plausible Compensation and Control of Photometric Variations in Radiance Field Reconstruction](https://mlsanity.com/paper/2601.18336/) (2026)
- [Efficient Camera Exposure Control for Visual Odometry via Deep Reinforcement Learning](https://mlsanity.com/paper/2408.17005/) (2024)
- [Learning Dense Feature Matching via Lifting Single 2D Image to 3D Space](https://mlsanity.com/paper/2507.00392/) (2025)
- [TaCOS: Task-Specific Camera Optimization with Simulation](https://mlsanity.com/paper/2404.11031/) (2024)
- [Exploiting Motion Prior for Accurate Pose Estimation of Dashboard Cameras](https://mlsanity.com/paper/2409.18673/) (2024)
- [Active Illumination Control in Low-Light Environments using NightHawk](https://mlsanity.com/paper/2506.06394/) (2025)
- [Reproducible Evaluation of Camera Auto-Exposure Methods in the Field: Platform, Benchmark and Lessons Learned](https://mlsanity.com/paper/2506.18844/) (2025)
- [A Feature-based Approach for the Recognition of Image Quality Degradation in Automotive Applications](https://mlsanity.com/paper/2303.07100/) (2023)
- [JOCA: Task-Driven Joint Optimisation of Camera Hardware and Adaptive Camera Control Algorithms](https://mlsanity.com/paper/2512.06763/) (2025)
- [Adaptive Illumination Control for Robot Perception](https://mlsanity.com/paper/2602.15900/) (2026)
- [Dual-Exposure Imaging with Events](https://mlsanity.com/paper/2604.10273/) (2026)

## Abstract (arXiv)

Successful visual navigation depends upon capturing images that contain sufficient useful information. In this letter, we explore a data-driven approach to account for environmental lighting changes, improving the quality of images for use in visual odometry (VO) or visual simultaneous localization and mapping (SLAM). We train a deep convolutional neural network model to predictively adjust camera gain and exposure time parameters such that consecutive images contain a maximal number of matchable features. The training process is fully self-supervised: our training signal is derived from an underlying VO or SLAM pipeline and, as a result, the model is optimized to perform well with that specific pipeline. We demonstrate through extensive real-world experiments that our network can anticipate and compensate for dramatic lighting changes (e.g., transitions into and out of road tunnels), maintaining a substantially higher number of inlier feature matches than competing camera parameter control algorithms.