AJOU Central Library Repository: Obstacle Detection on Roads based on Deep Learning

Graduate School of Ajou University Department of Mathematics 3. Theses(Master)

Obstacle Detection on Roads based on Deep Learning

Keyword: Complete Intersection over Union; Median Frequency Balancing; Obstacle Detection; VariFocal Loss; YOLOv7

Alternative Abstract: 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.

Appears in Collections:: Graduate School of Ajou University > Department of Mathematics > 3. Theses(Master)

qrcode

STATISTICS: Total Visit :5,012,794; Total Download :2,095; Today View :2,977

AJOU Central Library Repository는 국립중앙도서관 OAK 보급사업으로 구축되었습니다.