IASA Champ '24: UI Element Sketch Classification

Link to kaggle competition

• My first data science project and hackathon experience
• Timeline 27.01.2024-02.02.2024

Overview

The UI Element Sketch Classification Challenge involves classifying hand-drawn and AI-generated sketches of UI elements. The dataset comprises images from 21 distinct categories, such as alerts, buttons, and checkboxes. The task is to train a classification model capable of accurately identifying the class of a UI element in each image.

Dataset Description

• Train folder: Over 15,000 images are divided into folders corresponding to each of the 21 categories.
• Test folder: Includes images from all categories, without class indication.

Result

Building Sketch Classificator, based on Convolutional Neural Network (CNN) model

Dataset	Accuracy
Train set	0.848
Validation set	0.766
Test set	0.761