AMIEOD: Adaptive Multi-Experts Image Enhancement for Object Detection in Low-Illumination Scenes
Researchers have developed AMIEOD, a novel framework designed to enhance object detection in low-illumination environments. This system jointly optimizes image enhancement and object detection tasks, leveraging a Multi-Experts Image Enhancement Module (MEIEM) that employs various enhancement strategies. The framework also introduces a Detection-Guided Regression Loss (DGRL) and an Expert Selection Module (ESM) to dynamically adapt enhancement techniques based on detection performance. AI
IMPACT Improves object detection accuracy in challenging low-light conditions, potentially benefiting autonomous systems and surveillance.