This dataset had been originally shot for the ICPR 2012 human activities recognition and localization challenge (ICPR HARL). Target applications are the recognition of complex human activities, focusing on complex human behavior involving several people in the video at the same time, on actions involving several
interacting people and on human-object interactions.The dataset has been shot with two different cameras:
- a moving camera mounted on a mobile robot delivering grayscale videos in VGA resolution and depth images from a consumer depth camera (Primesense/MS Kinect); – a consumer camcorder delivering color videos in DVD resolution.
- The dataset comes with full annotation and several tools and consistsof the following parts:
- Video data (rgb, grayscale and depth frames).
- XML annotations of the activity bounding boxes.
- Calibration data and software for the MS Kinect depth sensor used during acquisition.
- Software for browsing videos and annotations (superimposed rectangles) and also for the creation of new annotations
- Software for evaluating new recognition software : automatic creation of precision/recall curves and integration of localization information into the evaluation process.
The dataset is centered on the following action classes:
- Discussion of two or several people.
- A person gives an item to a second person.
- An item is picked up or put down.
- A person enters or leaves an room.
- A person tries to enter a room unsuccessfully.
- A person unlocks a room and then enters it.
- A person leaves baggage unattended (drop and leave).
- Handshaking of two people.
- A person types on a keyboard.
- A person talks on a telephone.
LIRIS human activities dataset with depth data
Extracted from Imageworld: