Skip to main content

Face Recognition

Client:  A Law Enforcement Agency

Business Problem:  Recognition of suspects under surveillance in crowded public place

Data Challenges:   The suspect pictures are taken clandestinely from a distance by law enforcement agents with insufficient light, mix of crowds and varied backgrounds.

AI Solution: 

  1. Improve quality of suspects images taken from various angles before these images were used a training data to the Image recognition model. Various iterations were done to get decent quality upgradation of these images.
  2. Conduct a survey of the mall and recommend placement of cameras at the appropriate locations to capture best images of the people in the mall as this Camera network will be running parallel to the Mall’s own CCTV camera security system.
  3. We created lots of synthetic data from volunteers for the learning phase and also for the testing phase.
  4. We jointly worked with a large Image processing and recognition company to fine tune their image recognition engine to give best results for our clients data.
  5. We understood the ethical issues involved in False Positives. An innocent person identified as a possible suspect and may undergo hardship. We discussed this with the client and ensured human-in-the loop and transparent display of confidence levels from the AI engine.
  6. Proof-of-concept was completed in 4 months and limited live trial was implemented.

 

Solution Benefit:

  • As this was a confidential project, no details were provided.

 

Note:  These case studies are experiences and accomplishments by our Think-Tank members