Off-road Quad-Bike Detection Using CNN Models
| Autor: | Michael Abebe Berwo, Zhipeng Wang, Yong Fang, Jabar Mahmood, Nan Yang |
|---|---|
| Rok vydání: | 2022 |
| Předmět: | |
| Zdroj: | Journal of Physics: Conference Series. 2356:012026 |
| ISSN: | 1742-6596 1742-6588 |
| Popis: | Off-road vehicles are rapidly being employed for transportation, military activities, and sports racing. However, in monitoring and maintaining the race’s safety and reliability, quad-bike detection receives less attention than on-road vehicle recognition utilizing DL approaches. In this paper, we used transfer-learning approaches on pre-trained models of cutting-edge architectures, notably Yolov4, Yolov4-tiny, and Yolov5s, to detect quad-bikes from images and videos. A quad-bike dataset acquired from YouTube (https://youtu.be/ZyE3t3lG-vU. Accessed on April 10, 2022) was used to train and assess these designs. In this paper, we show that the Yolov4-tiny architecture outperforms the Yolov4, and Yolov5s in terms of mAP@50 and computing time per image. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|