Background fn

[adelchellabi]

Search before asking

• I have searched the YOLOv5 issues and discussions and found no similar questions.

Question

Hello
I don't understand what is background fn and background fp in confusion matrix (in the last row )
Please can you explain to me the meaning of this two things and thank you 😊

Additional

No response

[glenn-jocher]

@adelchellabi these are no different than TPs and FPs for any other class in the matrix.

[github-actions]

👋 Hello, this issue has been automatically marked as stale because it has not had recent activity. Please note it will be closed if no further activity occurs.

Access additional YOLOv5 🚀 resources:

• Wiki – https://github.com/ultralytics/yolov5/wiki
• Tutorials – https://docs.ultralytics.com/yolov5
• Docs – https://docs.ultralytics.com

Access additional Ultralytics ⚡ resources:

• Ultralytics HUB – https://ultralytics.com/hub
• Vision API – https://ultralytics.com/yolov5
• About Us – https://ultralytics.com/about
• Join Our Team – https://ultralytics.com/work
• Contact Us – https://ultralytics.com/contact

Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

Thank you for your contributions to YOLOv5 🚀 and Vision AI ⭐!

[hotohoto]

@glenn-jocher

So, you're saying background is another class? Then I believe the codes here are correct.

yolov5/utils/metrics.py

Lines 166 to 176 in d3ea0df

	for i, gc in enumerate(gt_classes):
	j = m0 == i
	if n and sum(j) == 1:
	self.matrix[detection_classes[m1[j]], gc] += 1 # correct
	else:
	self.matrix[self.nc, gc] += 1 # background FP
	if n:
	for i, dc in enumerate(detection_classes):
	if not any(m1 == i):
	self.matrix[dc, self.nc] += 1 # background FN

For each ground truth label, if there is no match, the model predicted it as background, so it's a false positive.

For each detection class, if there is no object hit, the model predicted it as non-background, so it's a false negative.

However, the visualization seems wrong. I mean Background FP and Background FN should be switched. Or we could drop only FP / FN. 🤔

[glenn-jocher]

@hotohoto True object (column) predicted as background (last row) is a FN, so labels are correct.

[hotohoto]

@glenn-jocher
Isn't L171 for the same case when true object gc is predicted as background? And the comment at the line says "background FP". As far as I understand the case will be visualized in the last row. So I feel like something is wrong.

Then do you think the comments in L171 (and L176) should be fixed? But if we say the background is a class, then we might be able to say the model predicted the area as positive of the background class but it is a false prediction, then we would call it a background false positive like the comment says. Personally, I like this explanation a bit more.

Anyway, if I got the code correctly, there seems to be a conflict between the comments in L171/L176 and the visualization. Still, it's possible I might have misunderstood the code though. 🤔

@rbavery

[glenn-jocher]

@hotohoto hi, thank you for your feature suggestion on how to improve YOLOv5 🚀!

The fastest and easiest way to incorporate your ideas into the official codebase is to submit a Pull Request (PR) implementing your idea, and if applicable providing before and after profiling/inference/training results to help us understand the improvement your feature provides. This allows us to directly see the changes in the code and to understand how they affect workflows and performance.

Please see our ✅ Contributing Guide to get started.