OptiCool-MachineLearning

👓 OptiCool 😎 - Stay in Style, See with Confidence (Team CH2-PS291) - ML Repository

Opticool Machine Learning Repository for Bangkit Capstone Project. Building face shape classifier Model.

🤓 The Coolers of Machine Learning Bangkit Academy Capstone Team CH2-PS291

Member	Student ID	Path	Role	Contacts
Rizky Intan Nurlita	M008BSX1236	Machine Learning	Machine Learning Engineer	rizkyintan
Sayyidan Muhamad Ikhsan	M008BSY1064	Machine Learning	Machine Learning Engineer	sayyidan-i
Ahmad Rosikh Al Muttaqy	M312BSY1903	Machine Learning	Machine Learning Engineer	creamysausage

Tech Stack

About

We have model, eyeglassess frame datasets, and recomendation algorithm in this repository

Face Shape Classifier

• Face Shape Classifier (image classification) use MobileNet as the base model for transfer learning that is taken from Keras. The model also contains an additional layer that received output from the based model. We use datasets Face Shape Image Dataset that contains 3,981 face images.

Model Training Performance

Performance after Fine Tuning

Performance Summary

Models	Accuracy	Val Accuracy
Face Shape Classifier	99.8 %	83.5 %

Scrapping for Eyeglasses frame from Online Marketplace

We have collected a dataset of 160 diverse eyeglasses frames from various online marketplaces using the web scraping process. The dataset, consisting of 160 entries, is intended to facilitate the development of an eyeglasses frame recommendation system based on classified face shapes. The eyeglasses frame dataset is stored in cleaned_data_final.csv

Metadata

• MarketplaceLink: Link to the eyeglasses frame product on the online marketplace.

• Name: Name or title of the eyeglasses frame.

• Brand: Brand of the eyeglasses frame.

• FaceShape: Classified face shape for personalized recommendations.

• Price: Price of the eyeglasses frame.

• Gender: Gender for which the eyeglasses frame is designed (e.g., Men, Women, Unisex).

• FrameColour: Color of the eyeglasses frame.

• FrameShape: Shape of the eyeglasses frame (e.g., Rectangular, Round, Aviator).

• FrameStyle: Style or design category of the eyeglasses frame.

• FrameMaterial: Material used to construct the eyeglasses frame.

• Pic1, Pic2, Pic3: Images of the eyeglasses frame from different angles.

• LinkPic1, LinkPic2, LinkPic3: Links or URLs to the images of the eyeglasses frame.

Each entry in the dataset includes these attributes, providing comprehensive information for eyeglasses frame recommendations based on face shapes.

Eyeglasses Recommendation Based on Face Shape

The primary purpose of the eyeglasses recommendation module is to enhance the user experience by offering curated suggestions that complement different face shapes.

Algorithm Recommendation

• The recommendation algorithm utilizes a combination of collaborative filtering and content-based filtering techniques to suggest eyeglasses frames. Collaborative filtering leverages user preferences and similarities, while content-based filtering considers the characteristics of eyeglasses frames and their alignment with face shapes.
• Users can explore recommended frames, view additional details, and provide feedback to further enhance subsequent recommendations.

Run the ipynb in Google Colab

You don't need to install anything just follow the steps below:

1. Download or clone this repository
2. Open Google Colab
3. Import the .ipynb file
4. Run the code

Run in Local

1. Download the .ipynb file or clone this repostitory
2. Run this locally using ex: jupyter notebook
3. Install all the dependencies

! pip install -r requirements.txt

4. Run all the code