Datasets
LIVIA provides specialized datasets to the community. These datasets serve as resources and references for the study, experimentation and advancement of image analysis and artificial intelligence techniques.
This page showcases the datasets collected at LIVIA for machine learning applications.
Behavioural Ambivalence/Hesitancy
The Behavioural Ambivalence/Hesitancy (BAH) dataset addresses a critical gap in emotion recognition research. Featuring 1,427 videos from 300 diverse Canadian participants, it captures subtle emotional conflicts through facial expressions, voice, and body language. With expert annotations, transcripts, and comprehensive metadata, BAH enables researchers to develop machine learning models for detecting ambivalence and hesitancy, complex emotions essential for personalizing digital behaviour change interventions.
SR-CACO-2
SR-CACO-2 fills a critical gap in microscopy imaging research. This dataset offers 2,200 images of Caco-2 cells across four resolution levels and three fluorescent markers, providing 9,937 training patches for super-resolution algorithms. By enabling researchers to enhance image quality without increasing harmful light exposure to living cells, SR-CACO-2 supports the development of advanced machine learning methods for biological imaging, with benchmarking results demonstrating substantial room for innovation.