Implement autodistill-yolov9

[Youho99]

Search before asking

• I have searched the Autodistill issues and found no similar feature requests.

Description

It might be great to implement yolov9 as a plugin in target models
https://github.com/WongKinYiu/yolov9

Use case

A simple detection model, like the tutorial with milk bottles

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[capjamesg]

Hello @Youho99! Thank you for creating this Issue. We would love to support YOLOv9 training in Autodistill. We would like to use the official repository, as we do in the Roboflow YOLOv9 training notebook. If you'd be interested in creating an Autodistill YOLOv9 repository, feel free to get started!

[Youho99]

The problem is that, apart from being constantly evolving, yolov9 is not packaged in pypi (and probably never will be)

I saw in your other implementations that you implemented the models present on pypi.

Do you have an idea on a development direction, for a non-packaged yolov9 model?

[capjamesg]

Here is an example of how to train a YOLOv9 model using a fork of the official repository that includes a few fixes: https://github.com/roboflow/notebooks/blob/main/notebooks/train-yolov9-object-detection-on-custom-dataset.ipynb

In cases where software is not distributed as a Python package, we typically add code to manually install the package and ensure it is read from the right path when Autodistill runs. See https://github.com/autodistill/autodistill-detic/blob/main/autodistill_detic/detic_model.py#L72 as an example of how we do this. It isn't ideal, but it works.