Problem

The project aimed to process a previously collected video database by blurring the faces of third parties and removing personal data from the recordings. This ensured that the data could be published ethically and in compliance with applicable law, supporting the development of transparent scientific research.

‍

Solution

We processed the recordings using an advanced algorithm published in 2022 at the prestigious CVPR conference. The anonymization process was carried out on a specially designed computer equipped with an NVIDIA graphics card, which enabled the processing of over 3 million frames in just two weeks. In the future, the use of the PL Grid infrastructure and the Cyfronet supercomputer is planned for more complex projects.

‍

Result

The anonymized dataset we created was published on the Kaggle platform, making it widely accessible to the scientific community. This contributes to transparent and reproducible research in the field of artificial intelligence. Several researchers have already begun working with the published data, highlighting its usefulness and value for scientific advancement.