MIDIH i-Deal-profile experiment

PROFILE
Personalized clothing design thROugh FashIon LifecyclE feeback loop
by MIDIH data at rest Knowage and Orion Context Broker tools
Speaker: Alessandro Canepa – i-Deal S.R.L.

Experiment Overview
PROFILE experiment aims demonstrating the potential for
design and manufacturing by exploiting Knowage tool and
Orion Context Broker to implement the industrial feedback
loop through the whole clothing product life
The experiment is related to Smart Product reference scenario.
Cross border experimentation will be carried out in
collaboration with VTT Technical Research Center for Real Time
Stream Data Analytics.

Experiment Overview
The simulations are based on 3 datasets, acquired by i-Deal:
• consumer morphology 3D measures and preferences
(acquired by ISizeYou app);
• clothing production measures and fitting trends (acquired
from the clothing manufacturers);
• e-commerce filtering and analysis (developed in Somatch
H2020 project).
The experiment enables this loop by providing 3 levels of
modeling and simulation:
• current design VS present consumers;
• new design VS present consumers;
• new design VS new target consumers;

Implementation
The components developed during the MIDIH experiment and are:
• Simulation APP: a web application to show datasets, launch simulations and show related results;
• Dataset sharing API: an API to share datasets from I-Deal to Orion Context Broker;
• Simulation API: an API to get data from Orion Context Broker, perform simulations and get back the results;
• Simulation result cockpit: a graphical interface to show simulation results by table and a set of charts, built directly on Knowage

Implementation
The components filled in blue are developed during the MIDIH experiment and are:
The Simulation APP is a graphical web interface (developed in PHP language) to drive the entire process from sharing datasets to
perform simulations and get back the results.
This web application drives the entire process to perform simulations. More precisely the workflow is the following:
• By using the web application, the operator shares the previous mentioned dataset with the Orion Context Broker by calling the
sharing API;
• Then, he/she launch simulations by calling the Simulation API;
• Finally, the simulation results are stored in Knowage database (useful for the cockpit) and retrieved by the web application

Implementation
The Login

Implementation
Dashboard to sum-up users' datasets of Simulation APP

Implementation
Population dataset of Simulation APP

Implementation
Clothing dataset of Simulation APP

Implementation
Simulation dataset of Simulation APP

Implementation
The Dataset sharing API is devoted to extract datasets information from local database (both for user and clothing) and to store them
into the Orion Context Broker. The API is developed in PHP language and is deployed within the Simulation APP.

Implementation
The Simulation API implements the core of the simulation activities. In particular, it offers the following functionalities:
• Read datasets information from Orion Context Broker and store into local Knowage database;
• Store simulation results to Knowage database;
• Extract simulation results from Knowage database to be exported into Orion Context Broker (optional) or to be accessible from
external service(s) or API(s);
• Perform classification of users and clothing;
• Perform simulation analysis.

Implementation
The Simulation result cockpit is implemented by using the cockpit functionality of Knowage tool, shows:
• A table that sum-up the simulation results;
• A chart to show ‘Current collection VS current population’ coverage results for male and female;
• A chart to show ‘Current collection VS new population’ coverage results for male and female;
• A chart to show ‘New collection VS current population’ coverage results for male and female;
• A chart to show ‘New collection VS new population’ coverage results for male and female.

Implementation
The Simulation result cockpit is implemented by using the cockpit functionality of Knowage tool, shows:
• A table that sum-up the simulation results;

Experiment Report
The objective of this experiment is to perform real-time simulations in order to detect the population coverage for a clothing design.
Coverage is expressed by a percentage that is required to:
• Help stylist during the definition of clothing measures and size development;
• Define the correct size development for clothing production phase.
During this experiment we consider the following populations:
• Current market (Italy): 4950 males and 4938 females;
• New market (Germany): 4948 males and 4934.
And the following collections for clothing:
• Current collection (Season 2018 of Piacenza): family 18 (fam18) for males and family 59 (fam59) for females, which respectively
contains #16 models for males and #9 models for females;
• New collection (Season 2019 of Piacenza): family 21 (fam21) for males and family 60 (fam60) for females, which respectively
contains #18 models for males and #11 models for females.

Experiment Report
The objective of this experiment is to perform real-time simulations in order to detect the population coverage for a clothing design.
Coverage is expressed by a percentage that is required to
• help stylist during the definition of clothing measures and size development;
• define the correct size development for clothing production phase.
Performing the four simulations on these datasets we obtain the following results:
Simulation no Simulation type Male percentage
coverage
Female percentage
coverage
1 Current collection VS
current population
57,66 % 54,46 %
2 Current collection VS
new population
36,76 % 30,52 %
3 New collection VS
current population
95,56 % 92,24 %
4 New collection VS new
population
75,02 % 84,86 %

KPIs collected
1. Design success rate: the sales season to retail is ongoing and the results are positive on the side of Piacenza, especially as regards
the fit feedbacks from foreign buyers. The retail F/W season will start in September and it will provide the final feedback from
consumers but, on the basis of the available information the expected value to increase from 5% to 15% will be reasonably
reached.
2. Design costs: the process applied to the families of products of F/F 2019/20 season of Piacenza has revealed an effective reduction
of the costs of product development due to the rapid real time feedback provided to the modellist as to support the new model
fitting definition. Extended to the whole clothing collection the process has the potential to reach more than 20% cost target
reduction. The aggregated impact at the whole collection level will result even increased when there is a prevalence of woman
designs, which have a higher rate of renewal in comparison with men’s ones, requiring a deeper design and fitting effort from the
style offices.
3. Design development timing: for F/W 2019/20 season of Piacenza the real time feedback as regards the fitting definition on the
basis of the simulations carried out for the new designs in relation with current and new target users has provided a direct and
effective support to the modellist in charge to define the measures of the new designs. The process has confirmed the potential to
reach the expected time reduction of 2-4 weeks of product development.

Exploitation
• PROFILE experiment has led to the implementation of a real time delivery of the 3 levels of simulation required to support the
design a new clothing collection on the basis of i-Deal services: current design VS present consumers (reference success rate of
present product), new design VS present consumers (simulated success of the new collection), new design VS new target
consumers (simulated success rate of new collection in a new market).
• The real time release of the results of them has removed a significant bottleneck to i-Deal service exploitation, which will start
from its established customers: Piacenza (active in the field of traditional clothing), Sparco (sport technical apparel) and Grassi
(worker protection clothing).
• Demonstrated the technical feasibility the additional efforts required will be focuses on the creation of the proper user interface,
in the first period dedicated to the internal operators of i-deal to better define and test them. In a second time they will be adapted
for the eventual direct use by customers designers

Conclusions
The objective of this experiment is divided into 3 phases:
• Dataset collection;
• Design and implementation;
• Run.
At the end of this phase we also perform tests on:
• Orion Context Broker APIs to get token, post, get and delete an entity
• Knowage to configure datasets, define cockpit tables and graphs

Technical conclusions
• During these tests we appreciate that Orion Context Broker is easy to adopt, in particular, that it is possible to use a public
instance. This simplifies the deployment because there is no need to install a dedicated instance of this component.
• On the contrary Knowage does not provide a public instance, therefore a fresh installation is needed. However, the setup of this
tool is quite easy.
• We encountered some difficulties on sharing a cockpit as a public URL but reading the documentation we understand that is
probably an issue related to the adopted version. Then our suggestion is to better explain and simplify the process to publish the
content of a cockpit.
• Orion Context Broker creates the opportunity to collect data from different sources: a) same provider and different sources
(Piacenza physical shops, not only e-commerce) b) different provider and different source: (sales from other vendors/designers and
from e-commerce and physical shops)
• Exploiting Knowage features, it is possible to elaborate more complex report, for example, adding dedicated functions to perform
economic impacts on sales and collections.
The obtained results demonstrated that experiment had success and it is possible to offer a business service to clothing designers in
order to reduce the collection failures.

Business conclusions
• It is advisable to concentrate the efforts on functionalities and data, relying on specific tools to collect (e.g. Orion Context Broker)
and report data (e.g. Knowage). Therefore, the lesson is trying to integrate existing component rather than create new ones. This
is very important for SMEs with low budgets;
• The design of new collection can get significant benefit from the deployed tool to help stylist in order to obtain high success rate
for a particular population and to reduce collection failures;
• The clothing measures and size development are strongly related to population morphotypes: to increase the success rate on a
new market, the morphotypes analysis and the matching simulation provides significant benefits.
In conclusion PRIVATE experiment has successfully demonstrated the possibility to run the 3 levels of simulation of its
service to support clothing design in real time by the exploitation of the potential of Orion Context Broker and of
Knowage, enabling i-Deal to provide its service in real time and removing a significant bottleneck to its commercial
development.

MIDIH i-Deal-profile experiment

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to MIDIH i-Deal-profile experiment

Similar to MIDIH i-Deal-profile experiment (20)

More from MIDIH_EU

More from MIDIH_EU (20)

Recently uploaded

Recently uploaded (20)

MIDIH i-Deal-profile experiment