Self-Stimulatory Behaviours in the Wild for Autism Diagnosis Dataset

Autism Spectrum Disorders (ASD), often referred to as autism, are neurological disorders characterised by deficits in cognitive skills, social and communicative behaviours. A common way of diagnosing ASD is by studying behavioural cues expressed by the children.

We introduce a new publicly available dataset (SSBD) of children videos exhibiting self-stimulatory (‘stimming’) behaviours commonly used in autism diagnosis. These videos, posted by parents/caregivers on public domain websites, are collected and annotated for the stimming behaviours.

These videos are extremely challenging for automatic behaviour analysis as they are recorded in uncontrolled natural settings. The dataset contains 75 videos with an average duration of 90 seconds per video, grouped under three categories of stimming behaviours:

  • arm flapping,
  • head banging, and
  • spinning.

We also provide baseline results of tests conducted on this dataset using a standard bag of words approach for human action recognition. The baseline approach uses a combination of STIP, HIG/HOF and Bag-of-Words, plus SVM for classification. The Matlab code provided in the dataset files below will download the already computed STIP files (about 4.5GB compressed for the 75 videos), which we provide.

Dataset Files: ssbd-release.zip (1.7MB)

SSBD_Licence – Licence file (PDF) – Make sure you read and agree to the conditions!

readme.txt – General description of the various files in the dataset – read this one first!

ssbd-paper-PID2968273.pdf – The ICCV 2013 workshop paper describing the dataset, annotation and baseline results (Download)

ssbd.m – Matlab script file to generate baseline results

url-list.pdf – The list of URLs for the actual video files (Note: The video files themselves are not distributed in the dataset. You will need to download them yourself.)

baseline-results.xlsx – Baseline classification performance

Annotations folder – Contains XML-based annotations for all videos in the dataset. The XML tag descriptions are given in the readme.txt file.

Note that you will also need to install the VLFeat library from http://www.vlfeat.org/.

Please contact Shyam Sundar Rajagopalan (Shyam[DOT]Rajagopalan[AT]canberra[DOT]edu[DOT]au) for any assistance and suggestions.