Upload model

After training a YOLOv5 or YOLOv8 model, you can upload the model to our platform for model-assisted labeling or to keep track of your different models. You can upload a model from a training run with the following command:

$ trainyolo project add_model <project_name> --type <model_type> --run_location <run_location> --run <run> --conf <conf> --iou <iou>

With:

• project_name: the name of the project

• model_type: either yolov5, yolov8, yolov8-seg or yolov8-pose

• run_location: location of the runs, default: ./runs

• run: name of the run, default: latest run

• conf: conf threshold, default: best f1 score

• iou: iou threshold, default: 0.45

This command will upload the best model (best.pt) from your latest run to your specified project, and set the confidence value to the one which maximizes the F1 score.

For example, to upload the latest training run from YOLOv8 to the project car detection, while keeping the default values for the optional arguments:

$ trainyolo project add_model "car detection" --type yolov8