How to get im_h_y from im_h?

[yuanshuai220]

I want to calculate PSNR on Y channel.
I have got im_gt_y and im_b_y through the mat format files.

im_gt_y = sio.loadmat("output/" + opt.image + ".mat")['im_gt']
im_b_y = sio.loadmat("output/" + opt.image + ".mat")['im_b']

im_gt_y = im_gt_y.astype(float)
im_b_y = im_b_y.astype(float)

psnr_bicubic = PSNR(im_gt_y, im_b_y, shave_border=opt.scale)

But I have some trouble in getting im_h_y from im_h

im_h_y = im_h[:,:,0].astype(float)
psnr_predicted = PSNR(im_gt_y, im_h_y, shave_border=opt.scale)

The result was not right.
I tried to transform the im_h to YCbCr by Image

im_h_ycbcr = np.array(Image.fromarray(im_h, "RGB").convert("YCbCr"))
im_h_y = im_h_ycbcr[0,:,:]
psnr_predicted = PSNR(im_gt_y, im_h_y, shave_border=opt.scale)

It still doesn't work.
Can you give me some advice here.

[twtygqyy]

@yuanshuai220 please also load im_h_y from mat file after converting rgb images into ycbcr images in matlab. PIL library cannot give you the best performance in calculating PSNR.

[MKFMIKU]

@twtygqyy can you provide the Matlab script to test psnr? 😂

[HPL123]

@yuanshuai220 ，Hi，I don't know why it needs so much memory when test and the train is normal. Have you ever had this problem?
::
RuntimeError: cuda runtime error (2) : out of memory at /opt/conda/conda-bld/pytorch_1501969512886/work/pytorch-0.1.12/torch/lib/THC/generic/THCStorage.cu:66