Video Analytics is the process of acquiring video data over streams, processing them, perform inference, extracting information, analyze it and take actions on the same. Essentially it converts pixel data to actionable insights. Video Analytics at the edge makes use of powerful embedded platforms placed at the camera source to perform all these operations locally and optionally transfer only the output to the cloud.
Embien has done numerous Edge Video analytics involving object detection and identification, object classification, tracking, counting etc. Our work on Video Analysis in Embedded Systems has been adopted in industries such as retail, entertainment, smart factory, transportation, livestock management etc.
Nvidia Jetson TX2, Nano, Xavier, Renesas RCar, NXP iMX8, Xilinx Kria SoMs etc.
Some of our algorithm’s expertise are CNN, SSD, YOLO, FasterRCNN, and MaskRCN
Intelligent video analytics - Nvidia DeepStream, TensorFlow, OpenCV, PyTorchVideo
Embedded Linux, Node JS, Angular, WiFi, Gigabit Ethernet, MIPI-CSI camera/ISP etc.
Embien’s Edge video analytics designs are highly optimized for embedded systems that low-latency and real-time response is achieved despite the inherent resource limitations. Our engineers can identify and fine tune the algorithms for specific business use cases and leverage classical approaches such as OpenCV processing as needed to take advantage of both the worlds.
Some of our recent works include people counting, facial recognition, license plate detection, parking occupancy, vehicle classifications and attendance tracking. We also have a proven Driver Monitoring System that uses various technologies to derive the alertness quotient of the driver. Our work on Industry 4.0 includes various UI based inspection system, quality assurance systems etc