Custom Training for YOLO Darknet Detector on Colab

So I used the notebook listed here: How to Train YOLOv4 on a Custom Dataset and followed the YouTube tutorial in conjunction with the article. However, when I reach the part where we actually start training the detector it gives me the error “/bin/bash: ./darknet: No such file or directory.”
Could anyone help me on what to do here?

Hi @Ali_Helmi, due to the change in the CUDA version and drivers in Colab, there are updates that need to be made to the cell where we configure the Makefile for building Darknet on Colab, and also the cell that ensures you have the correct architecture for your respective GPU.

I found a notebook and GitHub repo that claims to have fixed the issue. I’ll need to test it myself and reconfigure the structure so that it can be “run out of the box.”

However, if you’d like to start that process, you’re welcome to. I’m hoping to get to it today, and at worst by tomorrow. You’ll just be exporting in YOLO Darknet format, with show download code and pasting the snippet that appears.

Thank you so much @Mohamed for your help and support, I am eagerly waiting for your response.

I just tried the notebook you provided and followed it until the very end, however, when I reached this line “!./darknet detector train data/ cfg/yolov4-obj.cfg yolov4.conv.137 -dont_show -map.” It stopped working. I configured everything and all was working fine up to that point, and I even reworked the entire thing with the substitution of yolov4-tiny instead of the full version and that too worked until that line and stopped. Please advise @Mohamed!

Is there any other well documented notebook other than yolo darknet, which also features training on custom trained dataset, that is currently working?

We do have many more notebooks here:

The issue with YOLOv4 Darknet was added to the Issues in the repo, too.