the raw2rgb results have artifacts about cycleisp

[15757170756]

def pack_raw(im):
    img_shape = im.shape
    H = img_shape[0]
    W = img_shape[1]
    ## R G G B
    out = np.concatenate((im[0:H:2, 0:W:2, :],
                          im[0:H:2, 1:W:2, :],
                          im[1:H:2, 0:W:2, :],
                          im[1:H:2, 1:W:2, :]), axis=2)
    return out

[adityac8]

Hi
You need to clamp the output of the model_raw2rgb between 0 and 1.
Here is a snippet of code.

rgb_noisy = model_raw2rgb(raw_img_save, ccm_tensor)
rgb_noisy = torch.clamp(rgb_noisy, 0., 1.)
rgb_noisy = rgb_noisy.squeeze(0).permute(1, 2, 0)

Thanks