sid6i7 / execute

Fashion Recommendation System

A simple fashion recommendation system that can recommend the users items similar to something that they like. A user can input a clothing image and he will get several similar looking recommendations.

The tech-stack used:

Frontend: Flutter and dart
Backend: Django and python

Directory Structure

├── backend
│────├─ backend
│────├─ model
│────├─ manage.py
│
├── frontend
│
├── models
│────├── detector
│──────────└── dataset
│───────────────└── images
│───────────────└── styles.csv
│
├── config.py
│
├── controller.py

Steps

1. Data-Collection: A kaggle dataset was used, it consists of about 44000 images spanning across 140 categories of clothing.

2. Data-Preprocessing:

• Added a new column which consisted of image_path (pointed to the images folder) for each product_id.

• Images were converted into feature vectors using a pre-trained vgg16 model (with output layer removed)

3. Prediction:

• For a new input image, first it gets pre-processed and then cosine similarities are calculated for the new image feature vector and the remaining image vectors (that belong to the new image clothing category)

• N most similar items are the output