On the basis of YOLO, combined with multi-task learning to realize the tasks of traffic object detection, drivable area segmentation and lane line segmentation in unmanned driving, the unmanned panoramic perception algorithm based on DAFPN-YOLO is proposed. Then, dynamic attention is used to achieve scale perception, space perception and task perception to improve the model’s performance in three tasks: traffic object detection, driveable area segmentation and lane line segmentation. The multi-task loss function is adjusted, and FocalLoss is introduced to improve the model’s performance in the face of category-unbalanced data. The accuracy of vehicle perception algorithm for driverless cars is improved.
Improving Unmanned Panoramic Perception Algorithm of DAFPN-YOLO
Lect. Notes Electrical Eng.
Chinese Intelligent Systems Conference ; 2023 ; Ningbo, China October 14, 2023 - October 15, 2023
2023-10-08
14 pages
Article/Chapter (Book)
Electronic Resource
English
Autonomous Unmanned Vehicle Automatic Visual Tracking Based on SLAM and YOLO Algorithm
Springer Verlag | 2024
|