/embed-deepdream

tensorflowjs - deep dream

Primary LanguageJupyter NotebookMIT LicenseMIT

🐒 Deep Dream Monkeys

License: MIT

This is a TensorflowJS-Backed DeepDream sample code. Inspired by this kaggle notebook. Enjoy.

TODO:

  • Basic setup for npm and webpack
  • Image loading works
  • Image gradient can be obtained
  • Image update w.r.t. Mobilenet feature vectors
  • Update canvas and see results
  • Make an actual, working static website.
  • Fix visual artifacts for generated image
    • Increase model resolution
    • Replicate update loop from the original notebook
    • Make model update CI loop (use Kaggle?)
  • Add UI elements
    • Upload image capability
    • Various knobs and inputs (learning rate, steps, etc)
  • (Long term) Add GPU support

Installation

Clone this repo and npm install.

npm i

Usage

Development server

npm start

You can view the development server at localhost:8080.

Production build

npm run build

Note: Install http-server globally to deploy a simple server.

npm i -g http-server

You can view the deploy by creating a server in dist.

cd dist && http-server

Features

Dependencies

webpack

Babel

Loaders

Plugins

Linters

Authors

License

This project is open source and available under the MIT License.