Deep Learning Network for Object Detection Under the Poor Lighting Condition

Abstract

Object detection is a part of image processing that holds great importance in this modern world. To locate things in images or videos, a computer vision approach called object detection is utilized. To identify items in both still photos and moving films, numerous algorithms and models have been created. One of the main issues with such object detection algorithms is that they struggle to accurately identify things in images with poor lighting. This study tries to develop the best deep learning algorithm that can be used to predict items from photos with very little or no lighting to address this issue. For this,from GitHub, a dataset of images of various objects, such as tables, cats, and dogs, has been compiled. After categorization, the collected dataset is examined using two factors. The images are then preprocessed utilizing picture format conversion and histogram equalization. Two different deep learning models were produced by two different algorithms. The YOLO algorithm and the speedier RCNN algorithm are the two algorithms. The models are then trained and evaluated using the preprocessed dataset. The performance of the models during training and validation is assessed using a metric called the AP score. The YOLO algorithm has the better AP score overall, according to the analysis of the AP score. The YOLO is once again determined to be superior when the findings are displayed in a bar graph for easier understanding. The YOLO algorithm's prediction results are also examined, and it is discovered that even under conditions of extremely poor lighting, the algorithm can correctly predict the presence of one or more objects.