Skip to content

Use the MATLAB deep learning designer to modify the model structure #24

@listening510

Description

@listening510

We are currently training with the following code:
yolov8Det = trainYOLOv8ObjectDetector('data.yaml', 'yolov8n.pt', ImageSize=[720 720 3], MaxEpochs=10);
The structure of yolov8n.pt that we are using is officially designed. After running the code, we export it as an ONNX file using best.pt, which then generates a dlnetwork in the MATLAB workspace, and we can use the deep learning designer application to visualize the structure.
I would like to ask if I can modify the model structure in the MATLAB deep learning designer, add modules, and then let the code train on my newly designed dlnetwork structure instead of reading the structure from the pt file for training. It seems that the deep learning designer app can only export dlnetworks; can we train with this structure?
Currently, I am using MATLAB2024b, with Python 3.11 and Ultralytics downloaded as recommended by you. Thank you very much for your help and answers.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions