Obstacle Detection on Roads based on Deep Learning
- 주제(키워드) YOLOv7 , Obstacle Detection , Median Frequency Balancing , Complete Intersection over Union , VariFocal Loss
- 주제(DDC) 510
- 발행기관 아주대학교
- 지도교수 신동욱
- 발행년도 2023
- 학위수여년월 2023. 8
- 학위명 석사
- 학과 및 전공 일반대학원 수학과
- 실제URI http://www.dcollection.net/handler/ajou/000000032898
- 본문언어 영어
- 저작권 아주대학교 논문은 저작권에 의해 보호받습니다.
초록/요약
In this paper, we implement a deep learning model to detect obstacles on roads for the blind individuals and autonomous delivery robots. We applied various methods to get model with high accuracy. Firstly, we used YOLOv7-tiny as a deep learning model, which demonstrated excellent performance due to its efficient architecture. Secondly, we used median frequency balancing to solve class imbalance in the dataset, resulting in an increase of 0.4 in mAP. We also conducted experiments with different bounding box regression losses, such as GIoU, DIoU, and CIoU, as well as classification losses, such as FL, QFL, and VFL, to improve the efficient training and performance of the model. As a result, CIoU, which considers all important factors in bounding box regression, and VFL, which effectively addresses foreground-background class imbalance, showed the best performance with 68.7 mAP, surpassing the baseline by over 3% in mAP.
more목차
1 Introduction 1
2 Related Works 2
2.1 YOLO 3
2.2 Loss function for Bounding Box Regression 3
2.3 Loss function for Classification 4
3 Methods 5
3.1 Class Imbalance 5
3.2 YOLOv7 5
3.3 Loss function 8
4 Experiments 12
4.1 Dataset and Evaluation Metrics 13
4.2 Training Details 13
4.3 Experimental Results 15
4.4 Limitations 19
5 Conclusion 21
