Photography Science

World’s first initiative to match memory and scent data “Fragrant Memories : Flower Gift Project” – Send a flower bouquet with the scent of memories – Successfully achieved goal on crowdfunding in Japan in 5 days!

“Les mille feuilles de liberte” florists, located in 19 areas in the hot spots of Tokyo including Roppongi, has started the ”Fragrnt Memories : Flower Gift Project” an initiative which allows for the matching of the scents of flowers to the scent of memories.


Crowdfunding was setup to collect funds for the event on Japanese crowdfunding service “Makuake”, and the goal was achieved within 5 days. The event will be held at the Tokyo Marunouchi store between 17th until 22nd November.



Flowers are loved all over the world, as a ”standard” gift, however the annual amount of cut flowers per household is declining year by year. In Japan, the sales of the flower market as a whole is also gradually declining.


The project started with the aim of creating an opportunity for consumers who have not recently purchased flowers for gifts to easily find the right choice, within the flower market as a whole.


The project is a joint initiative in collaboration with the “Tokyo Metropolitan Central Wholesale Market”, who boasts the largest handling size of flowers in Japan and is the base line and indicator for prices of the flower market nation wide, aims to create a new idea around flowers for gifting.


Annual purchase of cut flowers per household

Source: Ministry of Agriculture, 2017, The present condition of flowers – Ministry of Internal Affairs and Communications statistics Bureau “Family Income and Expenditure Annual Report”


Challenging and creating a new idea around flowers for gifting

 Scent sensors will be developed for the project after the acquisition of funds via crowdfunding, which will be able to acquire detailed scent data, to match the scent of memories and flowers to provide a bouquet of flowers with the same scent. For example, for Japanese people, the scent of natto (fermented soybeans) increases your appetite, although is only a bad scent for people of other cultures. Whether people feel comfortable with a certain scent is strong related to the the past experiences which are recalled from the scent, rather than the strength of the scent itself.


However, scent measuring devices in the past measure simply the strength and weaknesses of scents and was not possible to judge whether the scent was ”pleasant”. This project is also an important attempt in such academic meaning.





The method of matching memory and scent

The device will extract digitized data, by analyzing scents from memories. Furthermore, using data the scents of areas as extra data, which is derived by the scenery within the photos of memories such as indoor/outdoor, by the water, in nature, in the city etc.


With the combined data, a match with the most similar scent of flower will be made, based on the accumulated data within the program inside the box, offering a unique bouquet of flowers with a memory behind it.



Examples of “Memorabilia” and “Photos of memories”


Honeymoon in Hawaii

Memorabilia – Ring, accessories Photograph – Photo taken at beach etc.


A new couple – Events/festivals

Memorabilia – Wrist band worn at the music festival Photograph – Photo taken at event or festival venue


The couple’s home

Memorabilia – Towel which the wife washes everyday Photograph – Pictures of the home


[Industry-academia collaboration that enables challenge]

In order to envision the realization of the creation of the flower gift project, the system was developed in collaboration with the foremost person in data analysis, Associate Professor Yu Ichifuji.



Yu Ichifuji

Nagasaki University

Center for Information and Communication Technology Associate Professor, Knowledge Platform Division Manager


Completed a of 3 year course, PhD in Information Science at Tohoku University Graduate School of Information Science and Electrical Engineering in 2010. In the same year, after working as a researcher at the Center for Integrated Research and Innovation of Information and Systems. In 2016, as an Associate Professor at Nagasaki University ICT Foundation Center, was engaged in research on data analysis on tourism. Received the 2015 DoCoMo – Mobile Science Award Social Science Department Encouragement Award.

Submitted a paper as a theme “Research and development of a recommendation system by scent data and machine learning” at IEEE BDTL 2017 (Big Data Transfer Learning) by The Institute of Electrical and Electronics Engineers, Inc. from December 2017 in Boston,U.S.A.

This project utilize that research contents as a mechanism of “matching of memory and scent” .


<Contents of Paper>

Research subject

Research and development of a recommendation system by scent data and machine learning


Background and objective

In e-commerce sites, a method of recommending products purchased by similar customers or items desired by the individual is deployed by utilizing user demographic data and purchase history data is becoming mainstream.

Meanwhile, it is difficult to create such systems and recommend in the same technique for goods are hard to distinguish, such as flowers and perfumes, because feeling “scent” differs among individual and scent is difficult to being clearly assigned to categories.


It is necessary to create new recommendation system for products which “scent” is an important factor in the product’s preference. Therefore, for this research, the target was placed on scents with unclear classification of categories to perform quantitative evaluation based on data obtained from sensors.


The development of a new recommendation system was developed by linking the data and colloquial expressions about scents and classifying them using machine learning, a model to recommend products from words and phrases reminiscent of the scents.


Research purpose

Development of a recommendation system and establishment of a new reasoning method from scent data and words/phrases express that.


Research content

  • Numerization and data collection of various scents using sensors
  • Collection and classification of words related to general scents
  • Linkage of the above two types of data using machine learning
  • Construction of a model for recommending optimal products from words/phrases based on scents
  • Development and implementation of recommendation system related to scent based on the reasoning model

*Scents: General smells which do not discomfort users


How it works

■  Fragrant Memories  Flower Gift Project – The Scent Sensor

A new device was developed for this project, combining a standard gas sensor as an scent sensor and a crystal oscillator which can measure a variety of scent substances as a molecular level.


■  Sensitive membrane utilizing a crystal oscillator

A “sensitive film generating device” was developed which allows scent components to adhere to the crystal oscillator as a thin film (sensitive membrane). Detailed scent data can be acquired by attaching 6 components to each quartz oscillator.



For inquiries

Kioku-no Kaori: Flower Gift Project PR Office: (Within Signal Co. Ltd.) Contact: Isogai, Futagami

TEL: 03-5572-6068 / FAX: 03-6673-4282 / E-mail: