A core part of my lab's philosophy is the commitment to reproducible research. We develop and release high-quality, fully-annotated datasets and open-source code to drive innovation and provide fair benchmarks for the entire research community.
Cardiac Acquisitions for Multi-structure Ultrasound Segmentation. A key benchmarking dataset for cardiac ultrasound analysis and method comparison.
The Automated Cardiac Diagnosis Challenge dataset, the reference benchmark for MRI cardiac segmentation software.
A large, fully-annotated brain dataset for training and testing deep learning models for brain fiber tracking (tractography).
MIOvision Traffic Camera Dataset. Dedicated to the localization and recognition of vehicles in real traffic images.
Dataset dedicated to the problem of motion detection via background subtraction methods.
Dataset dedicated to the problem of modeling the background image given a video with moving objects or persons in it.