ValueError: cannot reshape array

[yueqiw]

Hi,

I got the following error when running style transfer with segmentation masks.

Elapsed time in stylization: 0.003414
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
...
...
/home/yueqi/Dropbox/lib/FastPhotoStyle/process_stylization.pyc in stylization(p_wct, content_image_path, style_image_path, content_seg_path, style_seg_path, output_image_path, cuda)
     60 
     61     with Timer("Elapsed time in stylization: %f"):
---> 62         stylized_img = p_wct.transform(cont_img, styl_img, cont_seg, styl_seg)
     63     utils.save_image(stylized_img.data.cpu().float(), output_image_path, nrow=1)
     64 

/home/yueqi/Dropbox/lib/FastPhotoStyle/photo_wct.py in transform(self, cont_img, styl_img, cont_seg, styl_seg)
     33         sF4 = sF4.data.squeeze(0)
     34         cF4 = cF4.data.squeeze(0)
---> 35         csF4 = self.__feature_wct(cF4, sF4, cont_seg, styl_seg)
     36         Im4 = self.d4(csF4, cpool_idx, cpool1, cpool_idx2, cpool2, cpool_idx3, cpool3)
     37 

/home/yueqi/Dropbox/lib/FastPhotoStyle/photo_wct.py in __feature_wct(self, cont_feat, styl_feat, cont_seg, styl_seg)
     86                 if self.label_indicator[l] == 0:
     87                     continue
---> 88                 cont_mask = np.where(t_cont_seg.reshape(t_cont_seg.shape[0] * t_cont_seg.shape[1]) == l)
     89                 styl_mask = np.where(t_styl_seg.reshape(t_styl_seg.shape[0] * t_styl_seg.shape[1]) == l)
     90                 if cont_mask[0].size <= 0 or styl_mask[0].size <= 0:

ValueError: cannot reshape array of size 4128 into shape (1376,)

Fixing these two lines

t_cont_seg = np.asarray(Image.fromarray(cont_seg, mode='RGB').resize((cont_w, cont_h), Image.NEAREST))
t_styl_seg = np.asarray(Image.fromarray(styl_seg, mode='RGB').resize((styl_w, styl_h), Image.NEAREST))

to the following resolved the issue.

t_cont_seg = np.asarray(Image.fromarray(cont_seg).resize((cont_w, cont_h), Image.NEAREST))
t_styl_seg = np.asarray(Image.fromarray(styl_seg).resize((styl_w, styl_h), Image.NEAREST))

[wesleyw72]

Hi, This is also fixed in PR #33.

[yueqiw]

Thanks! didn't notice that before.