You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
hello,i tried to run the demo_view_synrthesis.sh with different images(not on the demo images),and the results were not satisfactory. And it seems that all the generated mesh are the same one. How can I inverse other images and get new mesh using other images?
The text was updated successfully, but these errors were encountered:
Hi, could you include the results here so I can help more? Also, since we adopt a GAN-based method, a good alignment is important. Aligning your input image to FFHQ landmarks is essential.
Hi, thanks for your reply. Visually inspected, the input image does not align with the FFHQ landmarks. Running this script with your image path as '--img_path' and then use the aligned(cropped) image as the input to this method, as you did now, is likely to give better performance. Feel free to let me know whether this could address the problem.
hello,i tried to run the demo_view_synrthesis.sh with different images(not on the demo images),and the results were not satisfactory. And it seems that all the generated mesh are the same one. How can I inverse other images and get new mesh using other images?
The text was updated successfully, but these errors were encountered: