Iranian Car Detection

Train

To train YOLOv5 with your custom dataset and use the model in Python scripts, follow these steps:

1. Install YOLOv5

git clone https://github.com/ultralytics/yolov5
cd yolov5
pip install -r requirements.txt

2. Prepare Dataset Structure

Ensure your dataset follows this structure:

yolov5/
├── data/
│   └── iranian-cars.yaml (your dataset config file)
├── datasets/
│   ├── train/
│   │   ├── images/
│   │   └── labels/
│   ├── valid/
│   │   ├── images/
│   │   └── labels/
│   └── test/
│       ├── images/
│       └── labels/

3. Train the Model

Run this command from the yolov5 directory:

python train.py --img 640 --batch 16 --epochs 100 --data ./data/iranian-cars.yaml --cfg models/yolov5s.yaml --weights yolov5s.pt --name iranian_cars --device 0

Parameters explanation:

• --img 640: Input image size
• --batch 16: Batch size (reduce if you get memory errors)
• --epochs 100: Number of training epochs
• --data: Path to your dataset config
• --weights: Pretrained weights
• --device 0: Use GPU 0 (remove for CPU)

4. Save the Trained Model

After training, the best model is automatically saved to: yolov5/runs/train/iranian_cars/weights/best.pt

5.Training Metrics Log

This document provides a log of the training process, including loss values, precision, recall, mean average precision (mAP), and learning rates across multiple epochs. The table below captures key metrics from epoch 0 to epoch 19.

Training Metrics Table

Epoch	Train Box Loss	Train Obj Loss	Train Cls Loss	Precision	Recall	[email protected]	[email protected]:0.95	Val Box Loss	Val Obj Loss	Val Cls Loss	LR0	LR1	LR2
0	0.029752	0.017216	0.044642	0.052204	0.99543	0.092053	0.069513	0.0074322	0.0025939	0.028605	0.07002	0.0033311	0.0033311
1	0.019484	0.011727	0.040809	0.36735	0.37564	0.23615	0.18769	0.0071434	0.0024565	0.027327	0.03969	0.0063346	0.0063346
2	0.017496	0.011097	0.037026	0.47352	0.42884	0.28394	0.2243	0.0057269	0.0024342	0.027143	0.00903	0.009008	0.009008
3	0.016027	0.011104	0.036067	0.4752	0.45217	0.40033	0.33215	0.0054337	0.0023839	0.026943	0.008515	0.008515	0.008515
4	0.014773	0.010564	0.032671	0.69246	0.45617	0.55299	0.46768	0.0050268	0.0021885	0.025733	0.008515	0.008515	0.008515
5	0.013968	0.010239	0.029778	0.75747	0.55689	0.64428	0.5471	0.0048111	0.0020525	0.026042	0.00802	0.00802	0.00802
6	0.01385	0.010062	0.028365	0.76277	0.56656	0.64012	0.55655	0.0044781	0.0021039	0.027025	0.007525	0.007525	0.007525
7	0.01322	0.0097443	0.026224	0.80079	0.59794	0.69828	0.61411	0.0045634	0.0019977	0.026165	0.00703	0.00703	0.00703
8	0.012651	0.009571	0.024953	0.79039	0.60317	0.70958	0.63084	0.0043376	0.0020838	0.025602	0.006535	0.006535	0.006535
9	0.01248	0.0094008	0.023718	0.8004	0.62057	0.72532	0.65015	0.0041919	0.0019356	0.025547	0.00604	0.00604	0.00604
10-19	(Additional epochs data to be filled)

6.Confusion Matrix

7. val batch

8. How To use:

from detection import detect_cars
detect_cars("detect_cars('/home/reza/Downloads/206.jpeg")

Fusing layers... YOLOv5s summary: 157 layers, 7045186 parameters, 0 gradients, 15.9 GFLOPs Adding AutoShape... Saved result to detected_206.jpg ('Peugeot-206', (14, 23, 555, 598)) '''shell