Ανίχνευση αντικειμένων για εφαρμογές αυτόνομης οδήγησης με χρήση μοντέλων YOLO και RCNN
View/
Keywords
Αυτόνομη οδήγησηAbstract
The main goal of this thesis is to identify objects for autonomous driving applications. To achieve this, we trained and evaluated two popular data detection architectures: the single-stage YOLO (You Only Look Once) detector and the two-stage Faster R-CNN detector.
The training and evaluation of our models was performed on the KITTI Vision Benchmark Suite, one of the most established datasets in the field of computer vision for autonomous driving applications. The dataset includes images of real road conditions with detailed annotations for objects such as vehicles and pedestrians, which were acquired in Karlsruhe, Germany.
Upon completion of the training and evaluation of the two models, we compare the performances achieved using metrics and emphasize the points that were achieved but also those that need improvement.