Datasets
SR-CACO-2: Confocal Fluorescence Microscopy Image Super-Resolution
NeurIPS 2024
A large-scale dataset for single-image super-resolution (SISR) in confocal fluorescence microscopy. Contains 2,200 unique images with four resolutions and three fluorescent markers, forming 9,937 patches for experiments. Includes real low-resolution images captured by microscope.
Key Features: Real LR/HR pairs, 3 upscaling levels (×2, ×4, ×8), 3 protein markers, ~16,800 multi-cellular objects
BAH: Behavioural Ambivalence/Hesitancy Recognition in Videos
2025
First multimodal dataset for recognizing ambivalence and hesitancy (A/H) in videos for behavioural change interventions. Collected from 224 participants across 9 Canadian provinces, containing 1,118 videos (8.26 hours total, 1.5 hours of A/H).
Key Features: Subject-based annotations, multimodal (video, audio, text), demographic diversity, A/H cues labeling, cropped faces, transcripts