Datathon24 - Kaggle Competition

Overview

This repository contains code and resources for the Datathon24 Kaggle competition. The challenge involves building a model to predict the design attributes of a product based on its image and metadata.

Pipeline

1. Feature Extraction: A pre-trained model is used to extract embeddings from product images.
2. Invalid Attribute Detection: A gradient boosting model identifies whether a product attribute is invalid.
3. Attribute Value Prediction: A neural network predicts attribute values for valid products.