3 ms·
I'm not sure, but there seems to be something wonky in the input images. They are very blocky, so I thought that they would be just pixel doubled (or quadrupled
by zokier 10y ago
I'm not sure, but there seems to be something wonky in the input images. They are very blocky, so I thought that they would be just pixel doubled (or quadrupled) from low-res pictures, but the blockiness lacks the regularity I'd expect from pixel-doubled images.
How were the input images prepared?
- anotheryou 10y agoSuper wonky indeed. Also it should compare to something like photoshops bicubic enlargement or the original size, because the brain gets stuck on the pixel edges.
- trustswz 10y agoIf you are interested in how it compares to bicubic or the original. Check these papers using the sub pixel convolutional layer: https://arxiv.org/abs/1609.05158 https://arxiv.org/abs/1609.05158 https://arxiv.org/abs/1609.04802 https://arxiv.org/abs/1609.04802.
- anotheryou 10y agothanks, impressive