Research Article
Precise Segmentation on Poly-Yolo, YoloV5 and FCN
@INPROCEEDINGS{10.1007/978-3-031-28990-3_5, author={Xinyuan Cai and Yangchenchen Jin and Yunfei Liao and Jiawen Tian and Yancong Deng}, title={Precise Segmentation on Poly-Yolo, YoloV5 and FCN}, proceedings={Edge Computing and IoT: Systems, Management and Security. Third EAI International Conference, ICECI 2022, Virtual Event, December 13-14, 2022, Proceedings}, proceedings_a={ICECI}, year={2023}, month={3}, keywords={}, doi={10.1007/978-3-031-28990-3_5} }- Xinyuan Cai
Yangchenchen Jin
Yunfei Liao
Jiawen Tian
Yancong Deng
Year: 2023
Precise Segmentation on Poly-Yolo, YoloV5 and FCN
ICECI
Springer
DOI: 10.1007/978-3-031-28990-3_5
Abstract
Nowadays, computer vision is becoming more and more popular and is applied to lots of fields. It can be used to detect safe and unsafe behavior happening in construction area [6]. It can also used for auto-vehicle driving, object detection, and so on. Currently, there are several ways to realize this like FCN and YOLO. However, they all exist some limitation. For example, the bounding box of YOLO is always being castigated by the user. There are a lot of versions of YOLO with different solutions to this kind of issues aiming to be used in different circumstances. Poly-YOLO is one of them. It decently solved the bounding box issue by using polygons rather than rectangles. In the paper, we focused on several techniques to achieve object detection. The YOLO approach will be mainly talked about for application, since it can be easily used combined with other algorithms like semantic segmentation, FCN, and etc. to enhance its performance. What’s more, the computation speed of Poly-YOLO is the main advantage why people choose YOLO.
