We present the Multiview Extended Video with Activities (MEVA) dataset, a new\nand very-large-scale dataset for human activity recognition. Existing security\ndatasets either focus on activity counts by aggregating public video\ndisseminated due to its content, which typically excludes same-scene background\nvideo, or they achieve persistence by observing public areas and thus cannot\ncontrol for activity content. Our dataset is over 9300 hours of untrimmed,\ncontinuous video, scripted to include diverse, simultaneous activities, along\nwith spontaneous background activity. We have annotated 144 hours for 37\nactivity types, marking bounding boxes of actors and props. Our collection\nobserved approximately 100 actors performing scripted scenarios and spontaneous\nbackground activity over a three-week period at an access-controlled venue,\ncollecting in multiple modalities with overlapping and non-overlapping indoor\nand outdoor viewpoints. The resulting data includes video from 38 RGB and\nthermal IR cameras, 42 hours of UAV footage, as well as GPS locations for the\nactors. 122 hours of annotation are sequestered in support of the NIST Activity\nin Extended Video (ActEV) challenge; the other 22 hours of annotation and the\ncorresponding video are available on our website, along with an additional 306\nhours of ground camera data, 4.6 hours of UAV data, and 9.6 hours of GPS logs.\nAdditional derived data includes camera models geo-registering the outdoor\ncameras and a dense 3D point cloud model of the outdoor scene. The data was\ncollected with IRB oversight and approval and released under a CC-BY-4.0\nlicense.\n