Neural network based image processing tool
Built with Electron and Python3.5
the communication part between electron and python is inspired by this repo
TL;DR:
A central demo platform for those lattest CV papers
Say you, a tech fan, have seen an amazing Computer Vision work, Laplacian Pyramid Super-Resolution, for example, via blog or news or forums or whatever. You think it's awesome and want to give it a try (use your image). Sadly, after some search you only find that you have to install matlab, matconvnet, download model and do some tricky command-line related work to get your result. And this is even not relatively painful compared with most CV work.
The real big problem is lots of CV paper authors not releasing their code or trained-model. There are, of course, many reasonable considerations. This phenomenon, which actually is common in code-related research fields, leads to a huge gap between the frontier and the public. There should be a bridge, and this tool is one step towards the goal, at least in the CV field.
Upsample selected images to specific resolution
Theano is used
Get the estimated depth map of a single chosen image, and feel free to add some built-in DOF effect
Tensorflow is used
cd to the project root directory in your terminal, then:
-
Make sure you have installed
nodejs
, version >= 7.4.0 -
(This is for mac users)
npm install fsevents -g
-
npm install [email protected]
-
npm install --runtime=electron --target=1.6.11
-
pip3 install -r requirements.txt
-
./download.sh
Now you can open the app via npm start