AutoRecycle is an intelligent recycling bin prototype designed to promote environmental sustainability through real-time waste classification. This system integrates machine learning and edge AI to classify waste effectively on a resource-constrained device like the Raspberry Pi 3B.
- Real-Time Classification: Classifies waste types using a camera and AI model.
- Edge Deployment: Optimized for deployment on low-resource devices.
- Interactive Design: Provides feedback via a display to educate users on proper recycling practices.
- Automated Segregation: Uses servo motors to direct waste to the appropriate compartment.
- Modular and Scalable: 3D-printed bin structure for ease of replication and customization.
- AI Model: Built using EfficientNet-B0 and fine-tuned for custom waste classification datasets.
- Hardware: Powered by a Raspberry Pi 3B with integrated distance sensors and servo motors.
- Optimization: Includes knowledge distillation to create a lightweight student model for efficient deployment.
- Public Dataset: Used for initial model training, sourced from Kaggle.
- Custom Dataset: Fine-tuned on real-world images for improved accuracy and relevance.
- Teacher Model: Achieved 90% accuracy on the test set.
- Student Model: Lightweight version with 88.5% accuracy for edge deployment.
- Inference Time: Approximately 5 seconds from detection to classification.
- Lighting Variations: Accuracy decreases in low-light conditions.
- Edge Cases: Misclassification of visually similar objects.
- Hardware Constraints: Limited by the computational power of Raspberry Pi 3B.
- Expand dataset to include more waste categories and scenarios.
- Explore advanced AI architectures for better generalization.
- Optimize hardware for reduced inference latency.
- Introduce additional features like waste volume prediction.
- Clone the Repository:
git clone https://github.com/your-repo-name/AutoRecycle.git