SmartStylist: A Fashion Recommender System powered by Computer Vision

Technical Features

• Object Detection Model: Leveraged the power of the YOLOv5 model trained on fashion images to detect fashion objects in images
• Feature Extraction: Utilized a Convolutional AutoEncoder implemented with PyTorch to extract latent features from detected fashion objects
• Similarity Search Index: Implemented FAISS library to construct an index, facilitating the search for visually similar outfits based on their distinct attributes

For more information on object detection model and feature extraction process, check out my repositories here:

• https://github.com/eyereece/yolo-object-detection-fashion
• https://github.com/eyereece/visual-search-with-image-embedding

Project Demo

Online Streamlit Demo:

Try the online streamlit demo.

Homepage:

Gallery:

Object Detection Model:

Getting Started

Clone the repository:

git clone https://github.com/eyereece/fashion-recommender-cv.git

Navigate to the project directory:

cd fashion-recommender-cv

Install dependencies:

pip install -r requirements.txt

Run the streamlit app:

streamlit run home.py

Usage

• Upload an image of an outfit (background in white works best)
• It currently only accepts jpg and png file
• Click "Show Recommendations" button to retrieve recommendations
• To update results, simply click on the "Show Recommendations" button again
• Navigate over to the sidebar, at the "gallery", to explore sample results

Dataset

The dataset used in this project is available here:

    @online{Eileen2020,
  author       = {Eileen Li, Eric Kim, Andrew Zhai, Josh Beal, Kunlong Gu},
  title        = {Bootstrapping Complete The Look at Pinterest},
  year         = {2020}
}