A vehicle's ability to run safely at high speeds requires the detection of objects accurately with real-time detection on the road. To certify a vehicle's safety at high speeds, visible objects on the road must be accurately detected in real-time. The proposed model is built using the YOLO v4 structure with the alteration in the backbone of the network. The backbone of the YOLO v4 model is CSPDarknet, which is replaced with CSPResNeXt for acquiring optimal speed and accuracy rate for detecting the object. The SPP and PAN together are taken as the neck and YOLO v3 is taken as the head of the network structure. The model has been developed with an alteration in the first part of the network, with CSPResNeXt in the YOLOv4 model, which does feature extraction and classification respectively. The model has been compared with existing models like Faster R-CNN, SSD and Mask R-CNN, and YOLO v2. Compared with these models, the proposed model provides optimal speed with better image resolution, high mAP values with less loss function
Alan : Eğitim Bilimleri
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|