Lero and UCD PhD student Abeba Birhane has won an award in the United States for a research paper which, while still under peer review, has already resulted in the withdrawal of an 80 million image dataset used to train AI. 

Abeba, along with Dr Vinay Prabhu, principal machine learning scientist at UnifyID, examined the problematic opacity, data collection ethics, labelling and classification, and consequences of large image data sets. The paper (under review) “Large image datasets: A pyrrhic win for computer vision?” is available at https://arxiv.org/pdf/2006.16923.pdf. Analysing the 80 Million Tiny Images dataset, the researchers found it contained thousands of images labelled with racist and misogynistic insults and derogatory terms.

Now, Abeba and Dr Prabhu have been presented with the Computer Vision Innovation award by leading tech news company VentureBeat at their second annual AI Innovation Awards. 

The winners are drawn from the website’s editorial coverage and the expertise of the nominating committee. The awards recognise people and companies making an impact in AI. The nominating committee includes Claire Delaunay, Vice President of Engineering, Nvidia and Asli Celikyilmaz, Principal Researcher, Microsoft Research. 

The other nominates in the computer vision category included Platform.ai, Dr Dhruv Batra, an assistant professor in the School of Interactive Computing at Georgia Tech and a research scientist at Facebook AI Research and robotic digitisation company Ripcord. 

Abeba is a PhD student supervised by Dr Anthony Ventresque of University College Dublin’s Complex Software Lab.