AI modules developed for surveillance cameras by Uncanny Vision
An Indian start-up is developing AI modules for CCTV cameras.
Uncanny Vision expects the innovation to help keep growing incidents of crime under check by predicatively warning about suspicious behaviour using deep learning algorithms. The Bangalore-based company has christened the solution Uncanny Surveillance. The software component enables it to analyse the attributes of people and objects with equal precision.
The founders of Uncanny Vision are exploring the opportunity of using IoT devices by optimising their software and embedding SoC (system on a chip) devices found in IoT hardware. Uncanny Surveillance can differentiate anomalous activity from normal patterns of activity and raise an alert to the control room. It promises to become more sophisticated at understanding patterns over time through deep learning.
[embedded content] Uncanny Vision aims to take image-analysis algorithms beyond Canny Edge Detection. The company made a research and development breakthrough with 'convolutional' neural networks (forms part of the tech universe that powers AI).
This raises the level of accuracy in image recognition to unparalleled heights thanks to higher GPU processing power and comprehensive image data availability (courtesy of cloud computing). Uncanny's founders are looking at using ARM processors, which have emerged parallel to the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) and house enough processing capability for advanced computer vision-centric tasks. Demand for Uncanny's solution is expected to come mainly from the retail industry, banking sector (for ATM booths), traffic management and facilities management, where the behaviour of people and attributes of objects make all the difference in predicting the possibility of a security breach.
The Chinese government is betting big on AI-based vision technologies. It has already committed huge investment to such technologies to enhance surveillance of its billion-plus population. Free download: The video surveillance report 2017
Sponsored by IDIS 'The Video Surveillance Report 2017' covers all things video surveillance based on a poll of hundreds of security professionals. Specifically looking at topics such as open platforms, 4K, low-light cameras, video analytics, warranties and this year due to the growing threat posed, the cybersecurity landscape. Click here to Download now
Umbo Computer Vision's SmartBullet camera strengthens intrusion detection with AI
Facial recognition needn't infringe our privacy as the CCTV commissioner fears.
Quite the opposite - if we get the tech right
A window to more than the soul?
AI research raises troubling questions about privacy, profiling and wrongful arrest