Artificial intelligence

Abstract: This work has conducted research and optimization of object detection methods in images using modern deep learning approaches. The work has conducted a theoretical analysis of the object detection problem, considered the role of computer vision in the modern information environment, and analyzed domestic and foreign scientific and technical sources. An analysis of existing neural network architectures, in particular YOLOv8, Faster R-CNN, and DETR, was conducted, with the determination of their advantages, disadvantages, and areas of effective application. The selected models were optimized by selecting hyperparameters, improving learning processes, and increasing the balance between accuracy and speed. Experimental implementation and comparison of models were conducted, which allowed assessing the impact of the applied optimization methods on the efficiency of detection systems.

Keywords: object detection methods, neural networks, computer vision, optimization, YOLOv8, Faster R-CNN, DETR

References:

1. Singh, P. (n.d.). Evolution of Object Detection: RCNN, Fast RCNN, and Faster RCNN. Medium. https://medium.com/@2003priyanshusingh/evolution-of-object-detection-rcnn-fast-rcnn-and-faster-rcnn-90cc872e6dae
2. S. Ren, K. He, R. Girshick and J. Sun, "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks," in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137-1149, 1 June 2017, doi: 10.1109/TPAMI.2016.2577031.
3. J. Redmon, S. Divvala, R. Girshick and A. Farhadi, "You Only Look Once: Unified, Real-Time Object Detection," 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 2016, pp. 779-788, doi: 10.1109/CVPR.2016.91.
4. Yaseen, M. (2024). What is YOLOv9: An in-depth exploration of the internal features of the next-generation object detector. arXiv preprint arXiv:2409.07813.
5. Ultralytics. (n.d.). Explore Ultralytics YOLOv8. Ultralytics YOLO Docs. https://docs.ultralytics.com/models/yolov8/
6. Carion, N., Massa, F., Synnaeve, G., Usunier, N., Kirillov, A., & Zagoruyko, S. (2020). End-to-end object detection with transformers. In European Conference on Computer Vision (ECCV) (pp. 213–229). Cham: Springer. https://doi.org/10.1007/978-3-030-58452-8_13
7. Jacob, B., Kligys, S., Chen, B., Zhu, M., Tang, M., Howard, A., Adam, H., & Kalenichenko, D. (2018). Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 2704–2713). https://doi.org/10.1109/CVPR.2018.00286
8. Han, S., Mao, H., & Dally, W. J. (2015). Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding. arXiv preprint arXiv:1510.00149.
9. Microsoft. (n.d.). Graph Optimizations in ONNX Runtime. ONNX Runtime Documentation. https://onnxruntime.ai/docs/performance/model-optimizations/graph-optimizations.html