Notable DETR-based object detection in low-light condition | |
| Author | Qi, Xiong |
| Call Number | AIT Thesis no.CS-24-03 |
| Subject(s) | Computer vision Optical detectors Optical data processing |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Computer Science |
| Publisher | Asian Institute of Technology |
| Abstract | Object detection often faces difficulties in low-light conditions due to noise and reduced contrast. While models like Retinex, SID, and KinD have advanced low-light image enhancement, integrating these improvements into object detection remains a challenge.This study proposes DualWaveNet, an efficient model combining Retinex theory and wavelet transform to enhance image brightness, contrast, and detail. Additionally, it introduces DW-CO-DETR, which integrates DualWaveNet with the CO-DETR frame work to boost object detection in low-light settings. The research focuses on two main questions: (1) How to design an effective low-light enhancement model? (2) How to integrate this model with high-performance object de tection frameworks? Through experiments on LOL-v1, LOL-v2, and ExDark datasets,DualWaveNet shows significant improvements in PSNR and SSIM. Furthermore, DW CO-DETR demonstrates notable gains in mAP and other detection metrics, surpassing the original CO-DETR in low-light environments. The results indicate that combining a robust low-light enhancement module with ad vanced detection models can effectively address low-light challenges, leading to better accuracy and reliability. This research provides valuable insights for future studies and practical applications in suboptimal lighting conditions. |
| Year | 2024 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Mongkol Ekpanyapong |
| Examination Committee(s) | Chaklam Silpasuwanchai;Huynh, Trung Luong |
| Scholarship Donor(s) | China Scholarship Council (CSC) |
| Degree | Thesis (M. Sc.) - Asian Institute of Technology, 2024 |