Click the link below to download the recyclable material dataset that we collected from five sources: TrashNet, Glassense-Vision, TrashBox, Waste Classification Data, and Waste_Pictures. The dataset consists of 12,873 RGB images, each of dimension 150x150 and labeled by one of 4 classes: glass, metal, paper, and plastic. The dataset distribution over the classes are shown in the following table.
| Label | Count | Percentage |
|---|---|---|
| Glass | 2,833 | 22.0% |
| Metal | 2,688 | 20.9% |
| Paper | 3,515 | 27.3% |
| Plastic | 3,837 | 29.8% |
Download link: Recyclable Material Data
For more details and classification results using CNN, refer to our paper Recycling Material Classification using Convolutional Neural Networks.