Dataset
Dataset Management
A high-quality, diverse dataset is crucial for training an accurate waste classification model. Our goal is to collect and label over 10,000 images of various waste items.
Image Collection & Labeling
Contribute to our dataset by uploading images of waste items. These images will be reviewed and labeled to improve our AI model.
Dataset Statistics
Current (Simulated) Dataset Size:
10,234 images
- Plastic: 3,500+ images
- Paper/Cardboard: 2,800+ images
- Organic: 2,000+ images
- Glass: 1,000+ images
- Metal: 500+ images
- Other: 400+ images
Continuous collection efforts are underway.
Model Retraining & Optimization
The ResNet model is periodically retrained with new data to improve accuracy and adapt to new waste item types. Optimization focuses on speed and power efficiency for embedded deployment on devices like the Nvidia Jetson Nano.
Developer Note
This section is a placeholder. Full model retraining pipelines are managed externally by the AI/ML team.