Making a license plate AI model Part 2

My previous attempt didn’t quite give me the results I was hoping for, so I looked for another way to train the model.

I still use Label Studio to make the dataset, the only thing that changed here is the export. I now use YOLO instead of Pascal.

These are my new files:

from ultralytics import YOLO
import datetime
import os

# Function to generate a timestamped name
def generate_experiment_name():
    return "train_" + datetime.datetime.now().strftime("%Y-%m-%d_%I:%M%p")

# Load a pretrained YOLO model (recommended for training)
model = YOLO('yolov8n.pt')

# Set project details
project_name = 'license_plate_project'
project_path = f'models/{project_name}'
current_experiment_name = generate_experiment_name()

# Train the model using the 'licensePlate.yaml' dataset for 400 epochs
results = model.train(
    data='licensePlate.yaml',
    epochs=800,
    resume=False,
    lr0=0.0004,
    project=project_path,
    name=current_experiment_name,
    batch=-1,  # Automatically set based on available resources in YOLO library.
    save_period=50,  # Save checkpoints every 50 steps (if supported by the library)
    degrees=180,
    shear=180,
    perspective=0.001,
    fliplr=0.5,
    cache=True,
    cos_lr=True,
    val=False,
    
)

# Export the final trained model to ONNX format and TFLite format
success_onnx = model.export(format='onnx')

# # Perform object detection on an image using the final trained model
# detection_results = model('../dataset/licensePlate/pascal_data/images/eed254b9-frame_1714924030.jpg', save=True)

import datetime
import os
from pathlib import Path
from ultralytics import YOLO

script_dir = Path(os.path.dirname(os.path.realpath(__file__)))
project_name = 'license_plate_project'
project_path = script_dir / 'models' / project_name

# Find the newest folder in models/license_plate_project/ that starts with a date
model_folders = [f for f in project_path.glob('*') if f.is_dir()]
newest_folder = max(model_folders, key=os.path.getctime)

# Load a pretrained YOLO model (recommended for training)
model = YOLO("/home/yuki/source/repos/license_plate_test/ultralytics/models/license_plate_project/train_2024-06-01_01:56PM2/" + 'weights/best.onnx')

# Perform object detection on an image using the model
results = model.predict('PARKEREN_Opladen_elektrische_autos_in_de_rij_2.jpg', save=True, project=project_path)

# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
path: /home/yuki/source/repos/license_plate_test/dataset/licensePlate/yolo/
train: images
val: images

# Classes
names:
  0: car
  1: licensePlate