For training, validation, and initial testing or RadCloud, we captured time-synchronized radar and lidar frames, sampled at a frame rate of 10 Hz, across 7 different environments including laboritories, corridors, and building lobbies. Our dataset comprises a total of 87,476 samples for training, validation, and testing. For the benefit of the research community, we make the dataset publicly available. For accessing the dataset and instructions on how to utilize the dataset, please reach out to David Hunt  (david.hunt@duke.edu)

Want to use our code or evaluate our models? Check out our github repository!