Litter Dataset

Is artificial intelligence capable of saving the Earth? We can’t say no if we haven’t tried. Litter is the number one cause of the environmental pollution. We’ve put a lot of effort into creating a proper taxonomy that can find its use in big industrial systems. We are very proud of our results and happy to share them with you!

Contact Us

Looking for a high-quality dataset?

Check out what’s already prepared, or send us an inquiry to arrange a custom dataset.


The Litter dataset contains 14k images with 20k annotations (bounding boxes) and 24 classes. Each class represents an object (cup), while subclasses determine its size, shape, or material (long paper cup/short paper cup). An Image Classification model is developed with 3004 images and its precision is 97%, while the Object Detection model has 2516 images and its average precision is 94%. All images were collected through our mobile application that is available both on Android and iOS and as such are property of General Blockchain, Inc. All annotations were created by our own fully employed in-house team. Each bounding cropped image was checked at least twice during the quality assurance process by different people.