Not Hotdog PWA

Hotdog Not Hotdog image classifier with TensorFlowJS. Demo link

Generated using [email protected]

Automated tasks

This project uses Gulp to run automated tasks for development and production builds. The tasks are as follows:

gulp --production: Same as gulp serve --production also run gulp test and not boot up production server

gulp serve: Compiles preprocessors and boots up development server gulp serve --open: Same as gulp serve but will also open up site/app in your default browser gulp serve --production: Same as gulp serve but will run all production tasks so you can view the site/app in it's final optimized form

gulp test: Lints all *.js file in the source folder using eslint and runs all *.test.js file unit tests through Karma and Jasmine gulp test --watch: Same as gulp test but will constantly watch *.test.js files and rerun tests when changes are detected

Adding the --debug option to any gulp task displays extra debugging information (ex. data being loaded into your templates)

Softwares

• Docker (https://store.docker.com/search?type=edition&offering=community): for initial setup of machine learning bits

Folder structure

training
├── Docker
└── data
    └── images
        ├── hotdog
        │   ├── hotdog1.jpg
        │   ├── hotdog2.jpg
        │   ├── hotdog3.jpg
        │   └── ...
        └── not-hotdog
            ├── not-hotdog1.jpg
            ├── not-hotdog2.jpg
            ├── not-hotdog3.jpg
            └── ...

Initial start up

Go to the root folder of the project and run the following code with Docker running in the background. Change /path/to/data to the full path of where the training images are. Take note that folders must not have spaces. It will cause training issues. The Docker file is included at the root folder of this project.

$ cd training
$ docker build -t model-builder .
$ docker run -v C:/xampp/htdocs/not-hotdog-pwa/ML/training/data:/data -it model-builder