Our AJILE dataset provides naturalistic long-term electrocorticography (ECoG) recordings with simultaneous upper body joint locations. To date, this is the largest and most comprehensive human dataset of its kind.

    We provide data from 4 subjects, each with about one week of continuous ECoG recording with 82 to 96 channels and joint locations from automated pose recognition. Because the joints are automatically detected, we also provide the confidence for each joint position. In total, at 30 frames per second, our dataset contains 7 (head, shoulders, elbows and wrists) joint locations for more than 60 million frames and approximately 560 hours of ECoG recordings. For more detailed information, see the readme page.

    If interested, please fill out the very short application link below. Our principle is that all legitimate research groups will have access to the data so only in extraordinary circumstances will an application be denied. The download link for the data should be available within a few days of the application.  

​

   Citation: Wang, NXR., Farhadi, A., Rao, R., & Brunton, B. (2018). AJILE Movement Prediction: Multimodal Deep Learning for Natural Human Neural Recordings and Video. In Proceedings of the 32nd AAAI Conference on Artificial Intelligence (AAAI).