Figure 1.
A sample picture of dataset.
Figure 2.
Image ANPR Flowchart.
Figure 3.
Examples of registration numbers.
Table 1.
YOLO license plate detection performance results.
| Model | Precision | Recall | F1-score | mAP@0.5 | mAP@0.5:0.95 | Inference Time | Epochs | parameters |
|---|---|---|---|---|---|---|---|---|
| YOLOv11n | 0.985 | 0.945 | 0.9645 | 0.963 | 0.692 | 1.7ms | 139 | 2,590,035 |
| YOLOv12n | 0.985 | 0.944 | 0.9640 | 0.963 | 0.689 | 1.9ms | 126 | 2,568,243 |
| YOLO26n | 0.979 | 0.937 | 0.9575 | 0.961 | 0.685 | 1.5ms | 94 | 2,504,190 |
Table 3.
Embedded detection performance of YOLOv11n, YOLOv12n, and YOLO26n on Raspberry Pi 5 and NVIDIA Jetson Nano.
| Metric | precision | recall | mAP@0.5 | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Raspberry Pi | NVIDIA Jetson | Raspberry Pi | NVIDIA Jetson | Raspberry Pi | NVIDIA Jetson | |||||||
| Model | PyTorch | ONNX | PyTorch | TensorRT | PyTorch | ONNX | PyTorch | TensorRT | PyTorch | ONNX | PyTorch | TensorRT |
| YOLOv11n | 0.969 | 0.946 | 0.942 | 0.924 | 0.671 | 0.702 | 0.697 | 0.771 | 0.741 | 0.746 | 0.741 | 0.828 |
| YOLOv12n | 0.939 | 0.916 | 0.962 | 0.940 | 0.745 | 0.700 | 0.723 | 0.681 | 0.780 | 0.753 | 0.753 | 0.737 |
| YOLO26n (PyTorch / TensorRT on Jetson) | 0.922 | 0.911 | 0.831 | 0.911 | 0.750 | 0.657 | 0.723 | 0.657 | 0.752 | 0.703 | 0.748 | 0.703 |
Figure 4.
ANPR on Raspberry Pi.
Table 4.
Inference time and throughput (FPS) of YOLO models on Raspberry Pi 5 and NVIDIA Jetson Nano using PyTorch, ONNX, and TensorRT.
| Metric | Inference time(ms) | |||
|---|---|---|---|---|
| Raspberry Pi | NVIDIA Jetson | |||
| Model | PyTorch | ONNX | Pytorch | TensorRT |
| YOLOv11n | 98.2 (10 FPS) | 38.3 (26 FPS) | 52.0 (19 FPS) | 38.9 (26 FPS) |
| YOLOv12n | 111.0 (9 FPS) | 40.7 (25 FPS) | 127.9 (8 FPS) | 88.2 (11 FPS) |
| YOLO26n (PyTorch / ONNX on Jetson) | 89.8 (11 FPS) | 33.4 (30 FPS) | 187.4 (5 FPS) | 137.6 (7 FPS) |
Figure 5.
Smart parking model for Smart City.
Table 5.
Smart contract Transaction cost.
| Smart Contract | Transaction Cost (×10−3 ETH) | Cost (USD) | Gas Units* |
|---|---|---|---|
| Access Control Smart Contract | 0.5 | 1.65 | 25,000 |
| Parking Fee Smart Contract | 0.8 | 2.64 | 40,000 |
| Vehicle Status and Data Integrity Smart Contract | 0.4 | 1.32 | 20,000 |
| Parking Slot Reservation Smart Contract | 1.0 | 3.30 | 50,000 |
1* Gas units were derived from the reported ETH cost and the fixed Ganache gas price of 20 Gwei (1 Gwei = 10−9 ETH) stated in Section 3.3, i.e., Gas units = ETH cost / 20×10−9. Each value in this table corresponds to the execution cost of a single invocation of the corresponding smart-contract function (not contract deployment, and not the aggregated cost of a full vehicle event, which may invoke more than one contract; see the discussion in Section 5).