Create new.py

new.py

@@ -0,0 +1,61 @@
+# Install required package
+# pip install ultralytics
+
+from ultralytics import YOLO
+
+# Dataset structure expected:
+# ├── dataset/
+# │   ├── train/
+# │   │   ├── images/
+# │   │   └── labels/
+# │   ├── valid/
+# │   │   ├── images/
+# │   │   └── labels/
+# │   └── test/
+# │       ├── images/
+# │       └── labels/
+
+# data.yaml example:
+# path: /path/to/dataset
+# train: train/images
+# val: valid/images
+# test: test/images
+# names:
+#   0: class1
+#   1: class2
+#   ...
+
+def train_yolov8():
+    # Load the YOLOv8 Large model
+    model = YOLO('yolov8l.pt')  # pretrained model
+
+    # Train the model
+    results = model.train(
+        data='data.yaml',
+        epochs=100,
+        batch=16,
+        imgsz=640,
+        device='0',  # 'cpu' or '0' for GPU
+        name='yolov8l_custom',
+        optimizer='Adam',
+        lr0=0.001,
+        warmup_epochs=3,
+        augment=True,
+        patience=50,
+        pretrained=True
+    )
+
+    # Validate the model
+    metrics = model.val()  # Validate on validation set
+    print(f"Validation [email protected]: {metrics.box.map}")
+
+    # Test the model (optional)
+    test_model = YOLO('runs/detect/yolov8l_custom/weights/best.pt')
+    test_metrics = test_model.val(data='data.yaml', split='test')
+    print(f"Test [email protected]: {test_metrics.box.map}")
+
+    # Export to ONNX format (optional)
+    model.export(format='onnx')
+
+if __name__ == '__main__':
+    train_yolov8()