Too Good To Be True: performance overestimation in (re)current practices for Human Activity Recognition

Open in new window