About TACO

TACO is a growing image dataset of waste in natural and urban environments. The dataset contains thousands of manually labeled litter photos from woods, roads, and beaches. Researchers and developers use TACO to train and evaluate object detection algorithms for automated trash identification.

Each image is segmented and annotated using a hierarchical taxonomy of waste categories. The structured labeling enables machine learning models to recognize specific litter types and locations. Furthermore, TACO’s open-source design allows the research community to contribute new images and improve annotation quality continuously.

Key Features

• Also includes diverse environmental contexts where litter accumulates and persists.
• Furthermore, images are manually labeled and segmented for high annotation accuracy.
• In addition, the dataset uses a hierarchical taxonomy to categorize waste types systematically.
• Additionally, the platform hosts images on Flickr and maintains an active collection server.

Access and Data

TACO is fully open-source and openly available for research, education, and development. Access the code and documentation on GitHub. The dataset continues growing through community contributions. Therefore, visit tacodataset.org to download images, submit new annotations, and join the collaborative effort to improve litter detection capabilities worldwide.