Train YOLOv9 Object Detection on a Custom Dataset | Step by step guide

Here is a short video tutorial on how to Train YOLOv9 on a custom dataset.
- 📷 YOLOv9 Blog Post: www.ikomia.ai/blog/train-yolo...
- 📓 Notebook: github.com/Ikomia-dev/noteboo...
- 📚Algorithm card (Ikomia HUB): app.ikomia.ai/hub/?q=yolo_v9
Resources:
- 🌏 Ikomia: www.ikomia.ai/
- 🚀 Ikomia API: github.com/Ikomia-dev/IkomiaApi
- 🌌 Ikomia Documentation: ikomia-dev.github.io/python-a...
- 📚 Ikomia Notebooks Repository: github.com/Ikomia-dev/noteboo...

[@Deepsim]

It is great. But how can we print the bbox and its center point?

[@ikomia7669]

Thanks Deepsim! To get the output values you can print 'yolov9.get_output(1).to_json()'. This will output a dictionary with the detection results. From that you calculate the center point. You can find more information on how to manage the Object Detection tasks following this link ikomia-dev.github.io/python-api-documentation/_autosummary/_autosummaryv/ikomia.dataprocess.pydataprocess.CObjectDetectionIO.html#cobjectdetectionio. If you need more freedom, the infer_yolo_v9 algorithm will be saved under 'user/Ikomia/Plugins/Python/infer_yolo_v9'. From here you have access to the process inference code and you can add prints if you want.

[@user-tx6ql4lz6s]

Hi ! You sound french haha, can you make a video on extraction of boxes with Yolov0 and SAM for segmentation ?

[@ikomia7669]

Hi ! Thanks for you comment. I will try to make a video on the subject but in the meantime you can use the notebook I just prepared. It shows how to create a workflow with YOLOv9 for object detection followed by SAM for segmentation. I hope this helps! github.com/Ikomia-dev/notebooks/blob/main/examples/HOWTO_use_YOLOv9_and_SAM_with_Ikomia_API.ipynb

[@tenns-tu1yz]

how to open code camera

[@ikomia7669]

Hi! Are is an example on how to run an object detection model on your camera: www.ikomia.ai/blog/yolov7-real-time-object-detection-with-ikomia-api

[@nassimnady3685]

It's possible to train yolov9 on multiple dataset ?

[@ikomia7669]

Hello ! Unfortunately it is not possible to train YOLOv9 on multiple datasets. You could try merging your datasets by using an annotation tool like CVAT or via scripts (some packages could be useful like COCO-merger).

[@nassimnady3685]

@@ikomia7669 thank's for your response. You know why I can't use the file "best.pt" with Ultralytics to show the prediction ?

[@ikomia7669]

​@@nassimnady3685 Ultralytics may utilize a variant of the model that's fits their API, potentially including alterations to metadata. I can see that there's a discrepancy in the size of the model weights between Ultralytics and the official repository. Without a detailed examination, it's challenging to specify the exact distinctions. The 'pre-trained.pt' file which available at the official repository (github.com/WongKinYiu/yolov9/releases/download/v0.1/yolov9-c.pt) might also not be directly compatible with Ultralytics.