search

javiribera / locating-objects-without-bboxes

PyTorch code for "Locating objects without bounding boxes" - Loss function and trained models
Other
251 stars 51 forks source link

validation time #35

Closed teeinn closed 4 years ago

teeinn commented 4 years ago

I know there's been many issues related with validation time. But to me, it's not clear.

I am training mall dataset and set --taus as default= -2 since it's faster than other ways. It still took about 4.5s per image for validation. I saw one posting that you said 'Only the neural network can use GPU. The thresholding and EM can only use CPU'. does that mean validation only has to be worked with CPU? Can I know the reason?

Thank you

teeinn commented 4 years ago

I close my case since I understood how this system works

VikasRajashekar commented 2 years ago

Can you please explain what happens in validation? I see that no GPU is used. @teeinn @javiribera

javiribera commented 2 years ago

During inference (either validation or testing), the probability map is converted to a binary mask using (1) static thresholding, (2) Otsu thresholding, or (3) a Beta Mixture Model-based thresholding, depending on the value of tau. Then Expectation Maximization is done to yield the final locations. This is all implemented in CPU.

javiribera commented 2 years ago

I know there's been many issues related with validation time. But to me, it's not clear.

I am training mall dataset and set --taus as default= -2 since it's faster than other ways. It still took about 4.5s per image for validation. I saw one posting that you said 'Only the neural network can use GPU. The thresholding and EM can only use CPU'. does that mean validation only has to be worked with CPU? Can I know the reason?

Thank you

I am surprised to hear this. The fastest way should by static thresholding (with a tau between 0 and 1).