atriumlts/subpixel
A TensorFlow implementation of the subpixel CNN architecture from Shi et al. (CVPR 2016) for single image and video super-resolution.
Not currently ranked — collecting fresh signals.
star history
This repository re-implements the Efficient Sub-Pixel Convolutional Neural Network paper by Shi et al. from Twitter VX. The network learns to upscale low-resolution images to higher resolution by using a novel sub-pixel convolution layer that reconstructs the output from feature maps at the final layer. Built with TensorFlow, it demonstrates the approach on image super-resolution tasks with example outputs showing progressive quality improvement over training epochs.
Frequently asked
- What is atriumlts/subpixel?
- A TensorFlow implementation of the subpixel CNN architecture from Shi et al. (CVPR 2016) for single image and video super-resolution.
- Is subpixel open source?
- Yes — atriumlts/subpixel is open source, released under the MIT license.
- What language is subpixel written in?
- atriumlts/subpixel is primarily written in Jupyter Notebook.
- How popular is subpixel?
- atriumlts/subpixel has 2.1k stars on GitHub.
- Where can I find subpixel?
- atriumlts/subpixel is on GitHub at https://github.com/atriumlts/subpixel.