# Exploring Low-light Object Detection Techniques

**Authors:** Winston Chen, Tejas Shah  
**arXiv:** [2107.14382](https://arxiv.org/abs/2107.14382)  
**Date:** 2021-07-30  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2107.14382v1/

## Key points

- Acquiring clear and accurate images under low light conditions is a significant challenge in computer vision.
- Low-light images exhibit undesirable characteristics such as high noise, poor illumination, reflectance, and low contrast.
- Extensive research has been conducted to enhance low-light images using pixel manipulation techniques and deep neural networks.
- Some approaches focus on improving illumination while others aim at reducing noise.
- Object detection neural network models have also been developed for low-light object detection.
- The study breaks down the problem into two phases: image enhancement algorithms and object detection models.
- Different image enhancement algorithms are explored, including histogram equalization and unpaired image translation techniques.
- Various object detection models are evaluated using mean average precision (mAP) calculations.
- The goal is to identify the most effective approach for low-light object detection through comprehensive analysis and experimentation.
- Insights are provided into optimal combinations of image enhancement algorithms and object detection models for detecting objects in low-light conditions.
- Potential directions for future research in this domain are discussed.

## Summary

In the field of computer vision, acquiring clear and accurate images under low light conditions poses a significant challenge. These images often exhibit multiple undesirable characteristics such as high noise, poor illumination, reflectance, and low contrast, which make object detection tasks particularly difficult. To address this issue, extensive research has been conducted to enhance these low-light images using various pixel manipulation techniques and deep neural networks. Some approaches have focused on improving the illumination in the images while others have aimed at reducing the noise. Additionally, considerable efforts have been dedicated to developing object detection neural network models. In this study, we aim to tackle the problem of low-light object detection by breaking it down into two distinct phases. In the first phase of our work, we explore different image enhancement algorithms to determine which one is better suited for object detection tasks where accurate feature retrieval takes precedence over image quality. Specifically, we investigate basic histogram equalization techniques and unpaired image translation techniques to enhance the low-light images. Moving on to the second phase of our work, we delve into exploring various object detection models that can be applied to the enhanced images obtained from the previous phase. By evaluating different models and comparing their performance using mean average precision (mAP) calculations, we aim to identify the most effective approach for low-light object detection. Through our comprehensive analysis and experimentation, we provide insights into which image enhancement algorithm and object detection model combination yields optimal results for detecting objects in low-light conditions. Furthermore, we discuss potential directions for future research in this domain. Overall, our study contributes to advancing the field of computer vision by addressing the challenges associated with low-light object detection. By refining image quality through enhancement algorithms and leveraging state-of-the-art object detection models, we aim to improve accuracy and reliability in real-world scenarios where lighting conditions may be suboptimal.