Interior Design Classification_project 5

Classifying Interior Design Styles Using CNN

By:Nadia & Manal

Project5

start

start

CONTENTS

Introdution

Data & Process

Model & Result

Evaluation & Conclusion

start

Introduction

Interior design involves a high amount of guessing. Although a room’s style can be predefined and categorized, these are typically hard to classify it by non experts. Thus, our goal is to classify some interior designs for different rooms based on their style; we focus on the two primary styles : modern, traditional (old). To achieve this goal, we utilized deep neural networks .

start

Tools

start

Data

Data source : kaggle [ https://www.kaggle.com/robinreni/house-rooms-image-dataset ] .

start

Process

Train the model transfer learning: 1- Feature extraction 2- fine Tunning

Result chose the best performance model

Preparing data " label dataset "

Preprocessing - Train-test split - image preprocessing: 1- Augmetation

start

Feature Extraction & Result

+ Baseline Model (EfficientNetB0): - Imagenet weights - epochs = 25 - learning rate = 0.001 - Batch size = 32 - Loss = 'binary_crossentropy' - Optimizer = ADAM

loss: 0.3902auc: 0.9226

start

Fine Tuning & Result

- Imagenet weights - epochs = 50 - learning rate = 0.000005 - Batch size = 32 - Loss = 'binary_crossentropy' - Optimizer = ADAM

loss: 0.0906auc: 0.9927

start

Evaluating the Model on New Images

start

Evaluating the Model on New Images

start

Conclusion

+ This model has 50 cycles, if we change the epoch to a big number we can get a more precise results

start

Thank you !

Any Questions ?