Welcome to Jetson AGX Orin repository

Model Training_Conversion

After completing the dataset annotation tutorial content, we can use the board to start training the model.

This tutorial only introduces CLI model training and conversion, you can refer to the official website for Python case modifications.

1、Model Training

Use CLI commands to directly train the model

xxxxxxxxxx
cd /home/jetson/ultralytics/ultralytics/data/yahboom_data/orange_data

xxxxxxxxxx
yolo detect train data=orange.yaml model=yolo26n.pt epochs=100 imgsz=640

data: Dataset configuration file

model: Pre-trained model file

epochs: Number of training rounds

imgsz: Specify input image size

Finally, the trained model-related files are stored in the /home/jetson/ultralytics/ultralytics/data/yahboom_data/orange_data/runs/detect/train folder.

Note: Each time training is run, the number after train will automatically increment by 1, please record this path, and use this new model path later.

2、Model Conversion

The final model will be generated in the runs folder: generally select the best.pt file for use

xxxxxxxxxx
/home/jetson/ultralytics/ultralytics/data/yahboom_data/orange_data/runs

Convert PyTorch format model to TensorRT:

xxxxxxxxxx
cd /home/jetson/ultralytics/ultralytics/data/yahboom_data/orange_data/runs/detect/train/weights
ls

xxxxxxxxxx
yolo export model=best.pt format=engine

Wait for the model conversion to complete, best.engine and best.onnx files will be generated in the current directory.

xxxxxxxxxx
ls

References

https://docs.ultralytics.com/modes/train/