ACADSTAFF UGM
| Title | : | Real Time Vehicle Make and Model System using Various of Yolo Architectures |
| Author | : |
Ika Candradewi, S.Si., M.Cs. (1) GILANG ARI SAPUTRA (2) |
| Date | : | 20 2021 |
| Keyword | : | vehicle recognition, make and model recognition, One stage detector vehicle recognition, make and model recognition, One stage detector |
| Abstract | : | In the current technological era, the Intelligent Transportation System (ITS) has begun to develop rapidly. One of the challenge of ITS is the vehicle make and model recognition system. The car model recognition system that is popular today is by using the ALPR (Automatic License Plate Recognition) system. The system will detect the license plate number and then access the database to get information about the car model. There are two drawbacks to this system. First, the system will not detect if the vehicle number plate is defective or not up to standard. The second disadvantage is that this system is running slowly and cannot recognize the car model in real time. Therefore, a car model recognition system is needed using a deep learning one stage detector method. In this study, there are four one-stage detector methods used, namely YOLOv3 5 layer detection, YOLOv3-SPP, YOLOv4, and Poly YOLO. Tests are carried out based on model evaluations during training and testing as well as measurement of computation time. In this test YOLOv3 5 layer got the best precision value of 93%, 95% recall, and 94?-Score. YOLOv3 5 layer can process with an average value of 21 frames per second |
| Group of Knowledge | : | |
| Level | : | Internasional |
| Status | : |
Submitted
|