Office of Commercialization  Semi-Automated Data Flow for Supervised Learning in Supervisory Review
Semi-Automated Data Flow for Supervised Learning in Supervisory Review

Semi-Automated Data Flow for Supervised Learning in Supervisory Review

Description:

Unmet Need: Fast and accurate identification of key performance indicators

Police departments intake huge amounts of visual data from body worn cameras. Processing video is time-intensive and can be difficult to interpret. However, analysis of this footage provides valuable information for officers and supervisors. Machine learning may be used to process and label visual data, unfortunately policing interactions are highly variable and not easily machine trainable.      

The Technology: Method of processing data using a blend of machine-learning and supervisor input

The inventors have developed software for semi-automated data processing that produces labeled data. The technology enables more efficient processing of data while creating critical labeled data for supervised learning. The semi-automated process allows for input and correction from officers and supervisors to continually improve the process.

Applications:

  • Efficient processing of policing data
  • Labeling data for supervised learning
  • Improved internal risk management

Advantages:

  • Integration of machine-learning and supervisor input
  • Semi-automated data flow
  • Labelling and processing large amounts of visual data
Patent Information: