About the demo.py inference.

[LWShowTime]

In the codes here, what is the function of the mask_transform? It seems like the h and w not to be the real size of the input image, otherwise we don't need to resize. I am confused with this.

Explicit-Visual-Prompt/demo.py

Lines 41 to 68 in bddc146

	h, w = list(map(int, args.resolution.split(',')))
	img_transform = transforms.Compose([
	transforms.Resize((w, h)),
	transforms.ToTensor(),
	transforms.Normalize(mean=[0.485, 0.456, 0.406],
	std=[0.229, 0.224, 0.225])
	])
	mask_transform = transforms.Compose([
	transforms.Resize((w, h)),
	transforms.ToTensor(),
	])
	inverse_transform = transforms.Compose([
	transforms.Normalize(mean=[0., 0., 0.],
	std=[1 / 0.229, 1 / 0.224, 1 / 0.225]),
	transforms.Normalize(mean=[-0.485, -0.456, -0.406],
	std=[1, 1, 1])
	])
	img = Image.open(args.input).convert('RGB')
	img = img_transform(img)
	img = img.cuda()
	pred = model.encoder.forward_dummy(img.unsqueeze(0))
	pred = torch.sigmoid(pred).view(1, w, h).cpu()
	transforms.ToPILImage()(pred).save(args.output)