Priberam

Next Seminar

Ethical Issues in Face Recognition: Explainable AI and Synthetic Datasets

Tue, 9 Apr 2024 13:00 - 14:00 WET

Abstract:

The outstanding performance gains in Face Recognition over the years have been sustained by three main factors. 1) Computational power that enabled deep neural networks; 2) Deeper and more complex network architectures; 3) Large datasets containing different identities and images. Due to these three points (and a few tweaks on the loss function) it was possible to surpass humans in this task. Yet, the latter two points brought concerning problems with them. The complexity of the architectures translated into black-box systems that cannot be fully understood nor explained. And the collection of the datasets has raised concerns regarding the collection method, consent and ethical use of that data. Additionally, it seems that models trained on these datasets are frequently unfair towards gender and ethnic groups. To mitigate these problems, we cover three topics: 1) How to adapt explainable artificial intelligence to face recognition?; 2) Why are my models unfair and what can I do about it? Is it all about balancing the number of samples per ethnicity?; 3) Are we already capable of creating fully synthetic datasets for face recognition? Do models perform well when trained in these datasets?. There are state-of-the-art approaches tackling each of these problems, from a new taxonomy for xAI to diffusion models and data balancing considerations (applied to our solution at the CVPR 2024 FRCSyn competition).

Register here on EventBrite.

The Priberam Machine Learning Lunch Seminars are a series of informal meetings which occur every two weeks at Instituto Superior Técnico, in Lisbon. It works as a discussion forum involving different research groups, from IST and elsewhere. Its participants are interested in areas such as (but not limited to): statistical machine learning, signal processing, pattern recognition, computer vision, natural language processing, computational biology, neural networks, control systems, reinforcement learning, or anything related (even if vaguely) with machine learning.

The seminars last for about one hour (including time for discussion and questions) and revolve around the general topic of Machine Learning. The speaker is a volunteer who decides the topic of his/her presentation. Past seminars have included presentations about state-of-the-art research, surveys and tutorials, practicing a conference talk, presenting a challenging problem and asking for help, and illustrating an interesting application of Machine Learning such as a prototype or finished product.

Presenters can have any background: undergrads, graduate students, academic researchers, company staff, etc. Anyone is welcome both to attend the seminar as well as to present it. Occasionally we will have invited speakers. Browse the archive (on the left) for a list of all past seminars, including the speakers, titles, abstracts and, whenever possible, the video and/or slides from the presentation.

Note: The seminars are held at lunch-time, and include delicious free food.

Feel free to join our mailing list, where seminar topics are announced beforehand. You may also visit the group webpage. Anyone can attend the seminars. If you would like to present something, please send us an email.

The seminars were usually held every other Tuesday, from 1 PM to 2 PM, at the IST campus in Alameda. This sometimes changes due to availability of the speakers, so check regularly! Since 2020, due to the pandemic, they have been taking place online, via Zoom.

The 2020-2021 season is now over. We will come back soon for a new season. Meanwhile please check some of the last seminars in Priberam’s YouTube channel.