Yolov5: Bug about check_anchors

[ChristopherSTAN]

In the train.py,

Line 202 should be:

check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)

[github-actions[bot]]

Hello @ChristopherSTAN, thank you for your interest in our work! Please visit our Custom Training Tutorial to get started, and see our Jupyter Notebook , Docker Image, and Google Cloud Quickstart Guide for example environments.

If this is a bug report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom model or data training question, please note that Ultralytics does not provide free personal support. As a leader in vision ML and AI, we do offer professional consulting, from simple expert advice up to delivery of fully customized, end-to-end production solutions for our clients, such as:

• Cloud-based AI systems operating on hundreds of HD video streams in realtime.
• Edge AI integrated into custom iOS and Android apps for realtime 30 FPS video inference.
• Custom data training, hyperparameter evolution, and model exportation to any destination.

For more information please visit https://www.ultralytics.com.

[glenn-jocher]

@ChristopherSTAN thank you noticing this and submitting a bug report! I just pushed a fix for this, should be resolved now. Please git pull and try again.

[ChristopherSTAN]

@glenn-jocher I feel excited about your Great job! I am training on VOC dataset, and I keep attention on yolov5, thanks again!

[glenn-jocher]

Great! Apologies for the bugs, lots of things are changing at the moment, but hopefully it will all smooth out a bit going forward.

[ChristopherSTAN]

Great! Apologies for the bugs, lots of things are changing at the moment, but hopefully it will all smooth out a bit going forward.

Another bug there.

If I use --evolve, code in line 403 of train.py:
# Evolve hyperparameters (optional) else: tb_writer = None
Will raise bug

[glenn-jocher]

@ChristopherSTAN hmm ok thanks! I will check it out.

[glenn-jocher]

Ok, got it fixed now!

[reactivetype]

@ChristopherSTAN did you need to create your own dataloader to work with VOC dataset?

[ChristopherSTAN]

@reactivetype No, just organize the images and labels as the tutorial showing.

[glenn-jocher]

@ChristopherSTAN I haven't trained on VOC myself, but it seems like a commonly used dataset. Perhaps we could add it to the /data folder by giving it it's own voc.yaml and an easy download command?

[ChristopherSTAN]

@glenn-jocher I am glad to. But I don't have much experiences of using git. So what should I do?

[glenn-jocher]

@ChristopherSTAN hmm, well a PR would be easiest (and then you show up as a repo author/contributor).

I can submit the changes also. All you need to do is upload your voc.yaml here, and some download commands, i.e. a download_voc.sh file also would help. Then I just throw them in the data folder. This might help a lot of people looking for voc training in the future.

[ChristopherSTAN]

OK, I will check the first way.
Wait a second.
Happy to be listed on your repo LOL.

[ChristopherSTAN]

@glenn-jocher I just did a PR.

I am not familiar with git operations. Sorry.

[glenn-jocher]

@ChristopherSTAN yeah I see it, thanks for submitting! It looks good, but I'm wondering if there's a way to reduce/merge some of the files. I will look it over more later today or tomorrow.

[ChristopherSTAN]

@glenn-jocher I think it could. I just copy them from my colab notbook. So they are kind of mess...

[ChristopherSTAN]

@glenn-jocher Besides, it is the reason why your work means a lot. That is, your model can easily trained on even colab.