Skip to content
A Demonstration of Deep Network Interpolation
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data
outputs
.gitignore
LICENSE
Pipfile
Pipfile.lock
README.md
main.py
model.py

README.md

dni-demo

Introduction

A demonstration of Deep Network Interpolation[1] for SR application. I use pretrained SRGAN/ESRGAN provided by MMSR[2].

Requirements

see Pipfile

Usage

python main.py --input path/to/your/image

Sample result

The interpolation result on Set14[3] baboon is here!

The Perception-Distortion Tradeoff[4] was observed.

Reference

  1. Wang, Xintao, et al. "Deep network interpolation for continuous imagery effect transition." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2019.
  2. MMSR from Multimedia Laboratory, CUHK.- https://github.com/open-mmlab/mmsr
  3. Bevilacqua, Marco, et al. "Low-complexity single-image super-resolution based on nonnegative neighbor embedding." (2012): 135-1.
  4. Blau, Yochai, and Tomer Michaeli. "The perception-distortion tradeoff." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018.
You can’t perform that action at this time.