Hand Gesture Retrieval in Art Image Collections (Master Project, Finished)


Manuel Rosenthaler


Hand gestures play a critical role in human communication from early times onwards. Understanding these hand gestures and analysing them in art collections of historical paintings can give an insight how this communication medium has changed throughout history. However, finding the instances of hand gestures in the large amount of art collections is a very cumbersome task and requires a lot of effort to analyze.

This project focuses on developing a hand gesture retrieval system in artistic image collections based on computer vision tools. Due to the different painting styles, the pipeline will require a hand detection which localizes the hands independent of the art style and a deep learning based representation learning method to discriminate the different hand shapes and gestures and retrieves the similar samples to the query image.

As an optional task, this project can be extended to vitrivr, by adding the feature extraction to the Cineast and using query by example to reduce the complications of query formulations for the search.

Start / End Dates

2021/04/01 - 2021/06/30


Research Topics