Skip to content

Latest commit

 

History

History
38 lines (28 loc) · 2.23 KB

README.md

File metadata and controls

38 lines (28 loc) · 2.23 KB

Find Me an Exhibit

Are you ready to take up the challenge? Film categories of objects in the exhibition "History of Switzerland" running against the clock.

The app displays one of several categories of exhibits that can be found in the exhibition (like "cloths", "paintings" or "clocks"). Your job is to find a matching exhibit as quick as possible. You don't have much time, so hurry up!

Best played on portable devices. ;-)

The frontend of the app is based on the game "Emoji Scavenger Hunt", the model is built with TensorFlow.js fed with a lot of images kindly provided by the National Museum Zurich. The app is in pre-alpha stage.

Demo

For a demo see here https://game.annotat.net

Set up your own environment

Requirements:

  1. git clone https://github.com/google/emoji-scavenger-hunt
  2. Go to the folder of the Dockerfile: cd emoji-scavenger-hunt/training/
  3. Build the docker image: docker build . -t model-builder
  4. Create a custom directory with a mandatory subfield images;
  5. Copy images to an arbitrary subfield of the images directory, each category of images in a dedicated field (e.g. for pictures on armours: /path/to/custom/dir/images/armours)
  6. Run the docker container to build the model: docker run -v /path/to/custom/directory:/data -it model-builder
  7. Copy the created models to the dist/model folder of the git repository: cp /path/to/data/saved_model_web/* dist/model
  8. Adapt the scavanger_class.ts in /path/to/custom/dir to your image categories (the name field of the objects must match to the names of your category folders)
  9. Copy the changed scavenger_class.ts file to src/js: cp /path/to/data/scavenger_class.ts src/js
  10. Install the needed dependencies: yarn prep
  11. Build the application yarn build
  12. Load dist/index.html in your preferred webbrowser