cig-skoltech / deep_demosaick Goto Github PK
View Code? Open in Web Editor NEWIterative Residual Network for Deep Joint Image Demosaicking and Denoising
Home Page: http://cig.skoltech.ru/deep_demosaick/
License: MIT License
Iterative Residual Network for Deep Joint Image Demosaicking and Denoising
Home Page: http://cig.skoltech.ru/deep_demosaick/
License: MIT License
Excuse me, I want to know how to train the denoiser before running the main.py to train the demosaicking model?
Thank you
Hi, this work is very interesting. When I testing the performance on my own captured image, I'm confused about the input shape of MMNet.
p = Demosaic(image_patch.float(), M_patch.float())
p.cuda_()
xcur = mmnet.forward_all_iter(p, max_iter=args.max_iter, init=args.init, noise_estimation=args.noise_estimation)
(1) In my testing, the size of image patch
is (H, W) with rggb bayer pattern. The largest number of image patch
should be 255.
(2) However, when I see the code in data_loader.py
, the size of input seems to have size of H, W, 3.
image_mosaic = np.zeros(image_gt.shape).astype(np.int32)
image_mosaic[:, :, 0] = mask[..., 0] * image_input
image_mosaic[:, :, 1] = mask[..., 1] * image_input
image_mosaic[:, :, 2] = mask[..., 2] * image_input
#print(image_mosaic.dtype)
image_input = np.sum(image_mosaic, axis=2, dtype='uint16')
# perform bilinear interpolation for bayer_rggb images
if self.apply_bilinear:
image_mosaic = self.preprocess(self.pattern, image_input)
image_gt = img_as_ubyte(image_gt)
image_input = image_mosaic.astype(np.float32)/65535*255
What's the right understanding of this question. Looking forward for your reply. Thank you.
Excuse me, your codes don't have the 'data' folder, so where and how can I download the dataset?
When running the test code written by myself, I can't obtain the results in your paper. Can you also release your test code? It would help a lot to check my mistake. Thank you
Excuse me, I want to know how to train the net-final.mat before running the main.py to train the demosaicking model? As if I don't load the resDNetPRelu_color_prox-stages:5-conv:5x5x3@64-res:3x3x64@64-std:[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]-solver:adam-jointTrain/net-final.mat before training the MSR dataset, the performance will drop dramatically. Could you please tell me how to train the net-final.mat
?
Thank you
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.