There seems to be a bug in the Mosaic data augmentation?

[haoyuefan]

I visualized the Mosaic data augmentation and found that some of the labels wrong shifted after the data augmentation. After debugging, I find there seems to be a problem with this line of the load_mosaic function:
x1b, y1b, x2b, y2b = w - (x2a - x1a), 0, max(xc, w), min(y2a - y1a, h)
the line 664 of dataset.py, I want to ask if this line should be changed:
x1b, y1b, x2b, y2b = w - (x2a - x1a), 0, w, min(y2a - y1a, h)
because in the bottom left patch of mosaic image, the xmax point of source image shouldn't be changed.
Hope you can reply me soon, thank you.

[github-actions]

Hello @yuefanhao, 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 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.

[gengyanlei]

@yuefanhao
I also found this problem, but it has no effect.

[haoyuefan]

@yuefanhao
I also found this problem, but it has no effect.

But it will create wrong label shift when I visualized the mosaic image with my own datasets, have you visualized the mosaic image?

[gengyanlei]

I used yolov5 to train my own data set and found no abnormalities. But when I studied mosaic code, I found this error by reasoning, but it didn't affect the image.
For example: a=np.ones([10,10,3]), a[:10] and a[:14] are equivalent.

[haoyuefan]

I used yolov5 to train my own data set and found no abnormalities. But when I studied mosaic code, I found this error by reasoning, but it didn't affect the image.
For example: a=np.ones([10,10,3]), a[:10] and a[:14] are equivalent.

ok, I understand that, thank you.

[glenn-jocher]

@gengyanlei @yuefanhao thanks for the bug report! Can you show an image of the mislabelled instances?

Would you like to submit a PR for this change?

[glenn-jocher]

@gengyanlei @yuefanhao I'm removing the TODO label and closing this issue, as this has been resolved in PR #1021 submitted by @yuefanhao and merged a few weeks ago. Thank you for your contributions!