reiinakano / fast-style-transfer-deeplearnjs Goto Github PK

Hey, I really like the project and would like to help with further improvements (e.g. adding the optimizations made by Dumoulin et al.), but it seems like this repository has exceeded it's LFS quota:

Error downloading object: src/ckpts/la_muse/Variable (5fb0a17): Smudge error: Error downloading src/ckpts/la_muse/Variable 
(5fb0a170666f9305d6b2640156b025af45a1c118cefb5da3708c1820a2613c7b): batch response: This repository is over its data quota. Purchase more data packs to restore access.

Is it really necessary to use LFS for the example images and checkpoints? As far as I know, a single checkpoint should weigh in at ~7MB, so I don't really see the need for LFS to store 42MB of checkpoints and a few images. I would create a pull request that adds the actual files, but I'm currently unable to clone the repository due to the error above.

Thanks for the great work!

Hello! I really liked Your project, it helped me a lot. I would like to ask you to port 4 new styles from this repository: https://github.com/hzy46/fast-neural-style-tensorflow. Everything on TensorFlow, and I think for You it will not be difficult. Unfortunately, I do not have the opportunity to do it myself.

Hi,

I really like your project.

Recently I tried it in the "watch" mode and decided to build it and then run on my local node.js server. After executing the deploy-demo script I got a single HTML file (730 kB) in a new folder (I called it "distr"). I tried to run it, but it does not work properly. It shows the controls, but does not show the pictures (source and style) and the button does nothing.

Please explain how to build the project properly and what files and folders does it need to work. For example, it requires a script from bower_components\webcomponentsjs folder, I copied this folder to the distr folder, but it changed nothing.

Thank you in advance.

Getting this error on Firefox 57 (beta) and Chrome 61 on Windows 10.

All of the requests seem to work properly, with adblockers disabled.

All I see is a white page w/ the github logo in the corner.

I went to https://reiinakano.github.io/fast-style-transfer-deeplearnjs/, but it didn't load at all. I checked the console, and there was a 404 on https://reiinakano.github.io/fast-style-transfer-deeplearnjs/bundle.js.

Hi

I really liked your work and I have tried on my local sytem and I want to know if I forked this repo and wanted to deploy on my github pages, then how do I do it ?

Google Chrome 61.0.3163.100 on debian buster/sid.

Hi, I clone your project, and run it as readme. But the demo is not working. When I clicked the transfer button, I got the follow error:
Uncaught SyntaxError: Unexpected token v in JSON at position 0.

Here is the return of network.:
version https://git-lfs.github.com/spec/v1
oid sha256:e03432142aa5e7f85776094d69bf5b90b11ba4498026d5ba2c73fefefbe35bf9
size 4689

I implemented Your work in C# for Windows users. You can find the code here: https://github.com/ColorfulSoft/StyleTransfer-Colorization-SuperResolution/tree/master/Style%20Transfer/2016.%20Perceptual%20Losses%20for%20Real-Time%20Style%20Transfer%20and%20Super-Resolution

Hello! I am pleased with Your project and would like to suggest you to implement another stylization algorithm proposed by Leon Gatis(A neural algorithm of artistic style). It allows you to stylize images for absolutely any style. There are a lot of realizations. Here's one made on Tenzorflow: https://github.com/woodrush/neural-art-tf. I hope this will interest you.

can this demo run on mobile browser?

When I build your project it writes that it is highly recommended to switch from Bower to Yarn:
https://bower.io/blog/2017/how-to-migrate-away-from-bower/

Are you going to switch it any time soon?

Thanks.

Error text ($ ./scripts/deploy-demo ... ) :

TypeScript error: src/net.ts(7,14): Error TS2420: Class 'TransformNet' incorrectly implements interface 'Model'.
Types of property 'predict' are incompatible.
Type '(preprocessedInput: Array3D<"float32" | "int32" | "bool">) => Promise<Array3D<"float32" | "int32"...' is not assignable to type '(input: NDArray<"float32" | "int32" | "bool", "0" | "1" | "2" | "3" | "4" | "higher">) => NDArray...'.
Type 'Promise<Array3D<"float32" | "int32" | "bool">>' is not assignable to type 'NDArray<"float32" | "int32" | "bool", "0" | "1" | "2" | "3" | "4" | "higher">'.
Property 'id' is missing in type 'Promise<Array3D<"float32" | "int32" | "bool">>'.

I found an implementation of a neural network to color images of the architecture you use: https://github.com/zeruniverse/neural-colorization. You can port it to DeepLearn.JS. The site will look 10 times easier. By the way, Your network implementation for image styling has helped me to make my implementation: https://github.com/PABCSoft/ImageStyler.