https://github.com/tarunrao0/lego-classification

A convolutional neural network (CNN) that classifies LEGO bricks into 16 different categories, enabling applications in automated sorting and inventory management
https://github.com/tarunrao0/lego-classification

computer-vision convolutional-neural-networks deep-learning image-classification

Last synced: 9 months ago
JSON representation

A convolutional neural network (CNN) that classifies LEGO bricks into 16 different categories, enabling applications in automated sorting and inventory management

• Host: GitHub
• URL: https://github.com/tarunrao0/lego-classification
• Owner: Tarunrao0
• Created: 2024-11-09T20:05:43.000Z (over 1 year ago)
• Default Branch: master
• Last Pushed: 2024-11-17T05:30:20.000Z (over 1 year ago)
• Last Synced: 2025-01-10T21:11:21.436Z (over 1 year ago)
• Topics: computer-vision, convolutional-neural-networks, deep-learning, image-classification
• Language: Jupyter Notebook
• Homepage:
• Size: 356 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          



    



# Lego Brick Classification with CNN

This project features a **Convolutional Neural Network (CNN)** that classifies LEGO bricks into **16 different categories**. The model is designed to facilitate applications in **automated sorting** and **inventory management**, making it easier for users to organize and utilize their LEGO collections effectively. By leveraging advanced image recognition techniques, this CNN not only enhances the sorting process but also opens up possibilities for more efficient inventory tracking and management in various contexts.

## Kaggle Dataset 

[Images of LEGO Bricks](https://www.kaggle.com/datasets/joosthazelzet/lego-brick-images)

A dataset containing 12,700 images of different lego bricks divided into 16 categories, each category containing 400 images

## Approach

- Normalized the images to have pixel values in the range[0, 1]

- Built a CNN with 4 convolutional layers, followed by pooling and dense layers

- Implemented Dropout layers to reduce overfitting

## Performance

- **Accuracy** : 95.0%

- **Precision** : 100.0%

- **Recall** : 99.8%

## Test Image Results

![Screenshot 2024-11-10 010230](https://github.com/user-attachments/assets/8b9aa16d-eadd-4465-ba38-264e24dc0f23)

![Screenshot 2024-11-10 013557](https://github.com/user-attachments/assets/ca5dff78-3021-4ad7-831c-a6190ccf0833)

# Links 🔗:

**Kaggle Dataset** : [Images of LEGO Bricks](https://www.kaggle.com/datasets/joosthazelzet/lego-brick-images) \

**Model with 95% accuracy** : [legos 95](https://drive.google.com/file/d/13LuV9sVNeApDJOtcoOrnc-8SDCnHwge_/view?usp=sharing)