Concepting
Prototyping
Developing
Entrepreneurship
Node.js
Python
OpenCV w/ Dlib
Machine Learning
2020
The Recnition Security system can use any camera stream as an video input to be processed for face detection and face recognition. Face detection is performed on live video streams in order to take pictures of faces automatically. These faces are stored and clustered per visitor for a maximum of 24 hours. The system can take up to 20-40 pictures of faces per second, depending on the hardware capabilities and amount of faces in the video.
Unwanted visitors can be added (within 24 hours) to a black list for a specific duration. If - for example - a shoplifter enters the store after being blacklisted 2 weeks ago, the system will recognise this person and notify the shop security within 3 seconds.
Faces are blurred due to privacy reasons.
The system provides real-time processed video streams to the Recnition App. Which makes it possible for security guards and other authorised personnel to see live video feeds including possible matches.
Faces of visitors are clustered (sorted) per visitor and saved for 24 hours. If a security guard wants to add a new (and violent) visitor to the blacklist, he/she can do so by clicking on that person in the visitors gallery. But what if there have been more than 2000 visitors in the past 24 hours? It then becomes very unpractical to scroll through the whole gallery.
This is where the search by image feature comes in handy. The Recnition App allows a security guard to select or take a picture of someone, and upload it for automatic face searching. After image processing the security guard will see a filtered gallery of potential visitors based on the person he/she uploaded and a certain percentage of face similarity.
Contact me!