/type-aware_model

Learning Type-Aware Embeddings for Fashion Compatibility Updated

Primary LanguageJupyter NotebookBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Evaluation and Comparison of Fashion Recommendation Models

This work is based on the PyTorch implementation for the Learning Type Aware Embeddings for Fashion Recommendation paper.

Usage

You can download the Polyvore Outfits dataset including the splits and questions for the compatibility and fill-in-the-blank tasks from here (6G).

This work presents the implementation of the following models:

  • Type-Aware Model
  • CSA-Net
  • CSA Fully-Connected Net
  • CSA Roberta Net
  • BYOL Dual-Net
  • Simple Siamese Dual-Net

Each model is available for both training and testing the results in a recommendation system. In other words, it returns a set of clothing items that are compatible with an input image.