CHIFFON is developed under the purpose of being a research platform for recipe displaying system driven by automatic scene recognition. This platform accelerate your research by skipping weirdly HCI interface implementation.

It consists of three modules, Viewer, Recognizer and Navigator. The recipe must be written in HWML.

The typical use case: a pattern-recognition researchers develop own recognizer for evaluating the effectiveness of the method to a real situation / A natural language processing researcher generate a machine readable recipe and test whether it is useful both for machine (pattern recognition system) and users, and so on.



The viewer provides a user interface for subjects. It shows directions of recipe step-by-step. The recipe must be written in HWML format. All the operation by subjects are once sent the navigator. The navigator consider both input from subjects and report from recognizer, and decide how to render the display.



The researchers freely develop recognizers. Only sending a HTTP get request enables the recognizer to control the viewer.

In the tutorial,  you can find the request patterns for controlling the viewer, through an interface for Wizard of OZ experiments.


The navigator is a back-end system. You normally do not understand how it works.  When you want to customize the rules ‘how triggers (in HWML) are fired by events,’ you can realize it by modifying this module. In that case, please contact us.


Click Here!


Any kind of proposal are welcome.


  title={How Does User's Access to Object Make HCI Smooth in Recipe Guidance?},
  author={Hashimoto, Atsushi and Inoue, Jin and Funatomi, Takuya and Minoh, Michihiko},
  booktitle={Proc. of 6th International Conf. on Cross-Cultural Design, Held as Part of HCI International 2014},

[:ja]Just another WordPress site[:en]- Kyoto University Smart Kitchen Dataset -[:]