thanks for the wonderful concept. can u please let me know exactly where in the co

adaptive cropping about imdn HOT 3 OPEN

zheng222 commented on September 26, 2024

adaptive cropping

from imdn.

Comments (3)

Zheng222 commented on September 26, 2024

@nan-rock

IMDN/test_IMDN_AS.py

Lines 30 to 60 in f1170cf

 def crop_forward(x, model, shave=32): 

 b, c, h, w = x.size() 

 h_half, w_half = h // 2, w // 2 

 h_size, w_size = h_half + shave - (h_half + shave) % 4, w_half + shave - (w_half + shave) % 4 

 inputlist = [ 

 x[:, :, 0:h_size, 0:w_size], 

 x[:, :, 0:h_size, (w - w_size):w], 

 x[:, :, (h - h_size):h, 0:w_size], 

 x[:, :, (h - h_size):h, (w - w_size):w]] 

 outputlist = [] 

 with torch.no_grad(): 

 input_batch = torch.cat(inputlist, dim=0) 

 output_batch = model(input_batch) 

 outputlist.extend(output_batch.chunk(4, dim=0)) 

 output = torch.zeros_like(x) 

 output[:, :, 0:h_half, 0:w_half] \ 

 = outputlist[0][:, :, 0:h_half, 0:w_half] 

 output[:, :, 0:h_half, w_half:w] \ 

 = outputlist[1][:, :, 0:h_half, (w_size - w + w_half):w_size] 

 output[:, :, h_half:h, 0:w_half] \ 

 = outputlist[2][:, :, (h_size - h + h_half):h_size, 0:w_half] 

 output[:, :, h_half:h, w_half:w] \ 

 = outputlist[3][:, :, (h_size - h + h_half):h_size, (w_size - w + w_half):w_size] 

 return output

from imdn.

nanmehta commented on September 26, 2024

thanks for the quick response.
i just wanted to ask as you mentioned in issue #9 regarding adaptive cropping tha it allows us to output exactly the same size as the target size. But does it make a difference in the overall result. As in the example you gave the output using this strategy is (2801x 2001) instead of 2800 X2000. But even we ignore 1 pixel from the boundary it wont make much difference

from imdn.

EvelyneCalista commented on September 26, 2024

Hello, may I ask if I would like to use ACS, shall I modify the train code with the crop_forward too?
thank you.

from imdn.

Recommend Projects

adaptive cropping about imdn HOT 3 OPEN

Comments (3)

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	def crop_forward(x, model, shave=32):
	b, c, h, w = x.size()
	h_half, w_half = h // 2, w // 2

	h_size, w_size = h_half + shave - (h_half + shave) % 4, w_half + shave - (w_half + shave) % 4

	inputlist = [
	x[:, :, 0:h_size, 0:w_size],
	x[:, :, 0:h_size, (w - w_size):w],
	x[:, :, (h - h_size):h, 0:w_size],
	x[:, :, (h - h_size):h, (w - w_size):w]]

	outputlist = []

	with torch.no_grad():
	input_batch = torch.cat(inputlist, dim=0)
	output_batch = model(input_batch)
	outputlist.extend(output_batch.chunk(4, dim=0))

	output = torch.zeros_like(x)

	output[:, :, 0:h_half, 0:w_half] \
	= outputlist[0][:, :, 0:h_half, 0:w_half]
	output[:, :, 0:h_half, w_half:w] \
	= outputlist[1][:, :, 0:h_half, (w_size - w + w_half):w_size]
	output[:, :, h_half:h, 0:w_half] \
	= outputlist[2][:, :, (h_size - h + h_half):h_size, 0:w_half]
	output[:, :, h_half:h, w_half:w] \
	= outputlist[3][:, :, (h_size - h + h_half):h_size, (w_size - w + w_half):w_size]

	return output