Describe your question/issue here! (delete this when you post)
I’m doing a YOLOv11 Object Detection (Fast) project.
On the “Model” page, there are a few training graphs.
I wonder how it comes up with these graphs. I can’t find any article about the formula of these graphs. I only know the brief meaning of them.
What do the numbers on the y-axis mean?
How are the numbers calculated?
Hey! A short answer is that the y-axis is the title of the chart. So for train/box_loss the loss starts at 1.5 and then improves (gets closer to zero) over the repeated epochs as the model learns and becomes more accurate. In this case, during training of the model, the accuracy of predicting a box around an object compared to what the data actually shows as the correct box (minimizing loss) improved over the 150 epochs.
There are also charts like val/box_loss which are the results of the trained model running against the validation dataset. You would expect that to improve as the model improves. (If not, you may have issues with how your training and validation images were determined - you want each to be a random, representative assortment. If one gets overloaded with a given scenario (like all red bikes) then the other will not reflect the same progress (if it has both red and yellow bikes for example.)
You can also look up “loss functions” if you are unfamiliar with that and want to better understand that piece of it.
If you have more questions be sure to ask as there are some great experts on here who can speak to this in detail!
This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.
