Accuracy Function

[PaperKites]

Hello

I'm wondering if you have used an accuracy function for determining the accuracy during the training because I couldn't find any metric function to tell how good is the output.

Also, does the code run over the whole dataset in each iteration (epoch) or does it pick 8 images randomly each iteration? (considering batch size=8)

Thank you in advance!

[naoto0804]

I haven't implemented but you can try some metrics like L1 or L2 error used in the following;
https://arxiv.org/pdf/1806.03589.pdf

You may refer to here for understanding how to load the data

[PaperKites]

So, Simply the code reads a random batch of images and gives it to the model every epoch, right?

pytorch-inpainting-with-partial-conv/train.py

Lines 107 to 111 in e04c84d

	for i in tqdm(range(start_iter, args.max_iter)):
	model.train()
	image, mask, gt = [x.to(device) for x in next(iterator_train)]
	output, _ = model(image, mask)