Addressing Research Gaps in Object Detection: A Model for Infrared-Based Low-Light Conditions

Object detection is the backbone of many modern real-world applications, including autonomous driving, surveillance, and robotics. As the object detection technology improves, particularly convolutional neural networks (CNN), the accuracy of such systems has improved drastically. Each of object have their own set unique characteristics which are used for purpose of object classification which is the backbone of object detection. Despite advancements there a few research gaps such as struggle to detect small and/or partially hidden objects, domain adaptation, multi-object tracking, unbalanced data, etc. Multiple approaches are used for object detection such as R-CNN, speedy R-CNN, Faster R-CNN, and YOLO to name a few. Finally, a model is proposed for object detection in low light conditions using near infrared imaging trained using YOLOv8 which can detect classes like vehicles and persons. The study resulted in maximum F1 score of 0.81 and mAP of 0.924.