Beta Presentation FIBRE: Fabric Identification Based Recommendation - - PowerPoint PPT Presentation

From Students… …to Professionals

The Capstone Experience

Beta Presentation

FIBRE: Fabric Identification Based Recommendation Engine

Department of Computer Science and Engineering Michigan State University Fall 2018

Team Herman Miller

Josh Bhattarai Ritwik Biswas Joe Smith Ted Stacy David Xuan

Project Overview

• FIBRE is an artificial intelligence fabric engine,

composed of a classification system and recommendation engine

• The classification component uses deep

learning and computer vision models to

• bjectively classify fabrics
• The recommendation engine correlates any

fabric a customer may submit with Herman Miller fabrics

The Capstone Experience Team Herman Miller Beta Presentation 2

System Architecture

The Capstone Experience Team Herman Miller Beta Presentation 3

Web-App Single Classification

The Capstone Experience Team Herman Miller Beta Presentation 4

Web-App Multiple Classifications

The Capstone Experience Team Herman Miller Beta Presentation 5

Web-App Recommendation

The Capstone Experience Team Herman Miller Beta Presentation 6

FIBRE Mobile App

The Capstone Experience 7 Team Herman Miller Beta Presentation

Autoencoder Based Recommendation

The Capstone Experience Team Herman Miller Beta Presentation 8

Recommendation Engine

The Capstone Experience Team Herman Miller Beta Presentation 9

Total Fabrics: 5,306

• Solid Texture: 3796
• Geometric: 476
• Stripe: 430
• Abstract: 422
• Botanical: 138
• Woven: 26
• Tufted: 10
• Felt: 8

Solid Texture 71.5% Geometric 8.9% Stripe 8.1% Abstract 7.9% Botanical 2.6% Woven 0.005% Tufted 0.002% Felt 0.001%

What’s left to do?

• Finish migrating Heroku app to AWS
• Add tokens for FIBRE API
• Improve recommendation engine algorithm
• Finalize iOS deployment (stretch)

The Capstone Experience Team Herman Miller Beta Presentation 10

Questions?

The Capstone Experience Team Herman Miller Beta Presentation 11

? ? ? ? ? ? ? ? ?