THI Labeling Tool

Annotated datasets are fundamental for pattern recognition and computer vision systems for at least two purposes: training of the algorithm and performance evaluation. In both cases, the quality of the annotation is important for successful development of the system, since patterns are found mainly on coherent data, i.e., the concept of "garbage in, garbage out" applies.

The THI Labeling Tool is an annotating software which contains several customized preferences and the automated data report generation. It accepts several different types of input data which embraces images, audio, point cloud, and general signals. As such, dealing with either indexed or sequence data is also foreseen. The output of our software is also customized. This open source framework is capable to handle and display multiple types of data, with the option of multiple users being able to remotely access the same database and work cooperatively.

Please follow this link to download the THI Labeling Tool.

If you use the tool, please mention publication credits: da Silva J.L., Tabata A.N., Broto L.C., Cocron M.P., Zimmer A., Brandmeier T. (2020) Open Source Multipurpose Multimedia Annotation Tool. In: Campilho A., Karray F., Wang Z. (eds) Image Analysis and Recognition. ICIAR 2020. Lecture Notes in Computer Science, vol 12131. Springer, Cham.

A presentation of the tool can also be downloaded.