This paper presents a comprehensive analysis of the evolution and current state of object detection algorithms in computer vision. The research focuses on three main frameworks: 1. Traditional detection algorithms 2. Anchor-based detection algorithms 3. Anchor-free detection algorithms The study highlights how deep learning has revolutionized object detection, moving from traditional methods to more advanced techniques based on deep neural networks (DNN). Key contributions include: * Detailed categorization and analysis of existing object detection frameworks * Comparison of two-stage and one-stage detectors * Analysis of anchor-based vs anchor-free detection methods * Experimental evaluation on common datasets * Discussion of current challenges and future research directions The research demonstrates how modern deep learning-based approaches have integrated feature extraction, selection, and classification into single models, achieving end-to-end performance optimization. This has led to significant improvements in both accuracy and efficiency compared to traditional detection schemes. The paper concludes with insights into potential future developments in the field of object detection, making it valuable for researchers and practitioners working in computer vision and deep learning applications.Recommended citation: Chi, B. (2022). "Research Advanced in the Object Detection Based on Deep Learning." 2022 International Conference on Applied Physics and Computing (ICAPC). DOI: 10.1109/ICAPC57304.2022.00092
Download Paper
