Is there a way to load yolov8 model?

[thloykot]

Hello! I have a pretrained yolov8 model and wonder if there any way to import it using kotlinDL?

[zaleslaw]

Do you have a pretrained YOLOv8 module in the ONNX format? If yes, please type here the ONNX inputs/outputs or other stuff related to the model specs. It's possible to import, I hope

Do you train it in the Python or C++? The main problem with YOLO is the porting of preprocessing/postprocessing code. We have no yet out the box, but if you share I could try to implement it with the KotlinDL

@thloykot

[thloykot]

Thaks for your reply @zaleslaw . Yes, have pretrained model. I trained it on Python, using ultralytics repository and then exported model to ONNX using their exporter. This model isn't final variant, so i can just share it with you. https://drive.google.com/file/d/1m_N7Eyd38HwsyGSE6Y0yn9hnTrh8D6nM/view?usp=sharing

[zaleslaw]

Thanks for sharing, as I understand you have your weights, but standard architecture, what was an initial point for you to take a basic model, could you please post a link

[thloykot]

I'm sorry, but i didn't dive so deep in this point, and don't thought about architecture