1 of 23

Style Busters

2 of 23

Logos

Clothes

Who Are We?

Artworks

3 of 23

Problem Statement

Online Shoppers and Fashion Designers suffer to find a similar style piece of clothes across millions of products on E-commerce websites.

Logo Generation Businesses Have No Way To Find Out If The Generated Logos Already Exists.

There Is Currently No Way To Find Out If The Style Of An Artwork Belongs To A Certain Artist.

4 of 23

Objectives

Easing The Process Of Finding Similar Style Clothes To Save Time and Effort by Automating This Process Using Deep Learning Techniques.

Preventing The Replication of Logos Caused By The Logos Generators.

Preventing The Stealing Of Artists Existing Artworks Style Using Deep Learning Techniques.

5 of 23

Clothes

System Overview

6 of 23

Artworks

System Overview

7 of 23

Logos

System Overview

8 of 23

Use-Cases

9 of 23

Functional Requirements

System shall be able to retrieve images of artworks with similar style to the uploaded image.

System shall be able to retrieve images of clothes with similar style to the uploaded image.

System shall be able to retrieve logos that are similar to the uploaded logo.

10 of 23

Non-functional requirements

The system should be maintainable and we will achieve this by documenting our system and using a clean architecture pattern when developing the system's mobile application.

The system shall be always up and running to receive requests from the users (internet connection is required to receive the results).

The system should be easy to use and this will be achieved by creating a user-friendly and organized user-interface.

11 of 23

Non-functional requirements

The system shall work on any Android or IOS powered devices.

The system must be secure and protected against any type of attacks and we will achieve this by taking various security measures such as encrypting the passwords of the users by hashing to keep their data secure.

The system should be reliable and we will achieve this by reducing down times by choosing a reliable hosting service.

Portability

Security

12 of 23

Class Diagram

13 of 23

Dataset

Our dataset is an extensive dataset of fashion designs, and it is called DeepFashion2 . Deep- fashion2 contains 292K diverse images from 13 different clothing categories that are famous for commercial shopping stores and consumers.

Source: https://github.com/switchablenorms/DeepFashion2

14 of 23

User Interface

15 of 23

How We Are Different

From

They Search By Look, We Search By Style.

1

Traditional Visual Search

Engines?

2 They Search Locally, We Search Remotely.

3 They Display Limited Results, We Will Display All The Results.

They Display Results From

Random Unverified Sources, We Will Display Results Only From Verified Sources.

4

16 of 23

Logos API

17 of 23

Logos Demo

18 of 23

Clothes Demo

19 of 23

Artworks API

20 of 23

Proposed Approach

Similar Look:

Similar Style:

21 of 23

Proposed Approach

VGG ARCHITECTURE

22 of 23

Our Team

Waleed

Abdelrahman

Sherif

Adham

Dr.Fatma

Eng. Mahmoud

23 of 23

Thank

you!