I am using SSD MobileNet V2 FPNLite 320x320 for training. I have got Total_loss graph for training step of 4000. Drawing graph for training loss is easy as it gives total loss for every iteration, e.g (from 100 steps to 4000 steps)
But I want to draw graph for validation loss which is 0.26. In validation loss it just gives the total loss. How should I draw my line graph for validation loss as I have just one point. There is a paper that have drawn graph for their validation loss but I don’t know how. Please I want the same curly graph for my model.
Hi @Annie,
If i understand correctly, you are not recording validation loss every epoch or every few epochs That is why you are seeing one final validation loss point.
To draw a graph for validation loss similar to the one you have for training loss, you need to calculate the validation loss at multiple points during the training process. This involves evaluating your model on a validation dataset at regular intervals and recording the validation loss at each of these points.
I just created the sample gist on how to plot the graph for validation loss at each interval. Which might help you.
Thanks.