Itamar Roth has over 15 years of start-up experience in business, marketing and product development with an in-depth focus on embedded systems and imaging solutions.
Prior to PointGrab, he led marketing and product management for the enterprise division at Anobit (acquired by Apple) and managed the professional services both at TransChip (acquired by Samsung) and at the Samsung Israel R&D Centre. Here he answers some much asked building management questions.
Can you describe your products and services and highlight what differentiates you from other solutions?
Our solution, CogniPoint, is a miniature network-connected computer vision sensor, running embedded analytics on a low-cost processor. This ceiling-mounted device comes in two form factors: incorporated into lighting fixtures or installed as a stand-alone node connected within a building network. The sensor leverages the advantages of the vision modality to capture fine object and occupant data, but because all analytics are performed at the sensor level, images are never transmitted from the sensor, only the processed data or actionable information is delivered as the output. The sensors adhere to all communication standards and have powerful security mechanisms built directly into the design. CogniPoint, which leverages deep learning algorithms, is unique in the vast amount and detailed richness of the data provided, without ever compromising the occupants’ privacy.
What do you see as the main value of your product?
Since the intelligence of a building depends on the sensor data available to it, the value of our sensor is substantial across all building functions. In addition to significant energy savings, improved safety and enhanced security, we see that accurate occupant activity and location data brings unprecedented value for optimizing space utilization. This is reflected in numerous applications such as hot desking, space planning, facility cleaning, room allocation and more.
What are the alternative solutions that currently address building automation needs?
There are a variety of relevant sensors available, but they address only certain aspects of building automation functionality. These include motion detectors, photocells, Real Time Location Systems (RTLS) and surveillance cameras. Motion detectors excel at detecting human presence but lack the resolution and capability to count and track occupants. Photocells are useful in detecting light levels for daylight harvesting but are limited in detecting actual light distribution in a room. RTLS technology tracks wireless devices, such as smartphones, wearables and RFID tags, but you must wear one at all times in order for these to be effective. In addition, RTLS accuracy is limited and provides only location information. Surveillance cameras are capable of the same functionality as CogniPoint sensing solutions, but these cameras are significantly more expensive and their video streaming capability is a heavy load on network traffic. Most importantly, surveillance cameras violate occupants’ privacy, a sensitive issue in every building environment.
What segments in the commercial building market are you targeting?
Our focus is on office buildings where we are collaborating with both large integrators and leading lighting OEMs. The impact of the Internet of Things (IoT) revolution on office building automation is apparent and traditional players now acknowledge the importance of quality sensor data in this market segment. We also pursue opportunities in the retail space, together with solution providers, and see healthcare as a particularly interesting and rapidly developing vertical.
Can you give us some insights into the underlying technology of the CogniPoint sensing solution?
To address the challenge of providing embedded-analytics sensors we adopted deep learning technology, an emerging state-of-the-art technology in data-driven machine learning systems. With cameras embedded at the sensing level, deep learning algorithms are able to perform end-to-end computation -- from raw sensor data all the way to the final output. The scalability and flexibility of deep learning technology makes it a powerful capability for real-time systems such as a smart sensor in the challenging environment of commercial buildings. Moreover, the sophisticated fusion of deep learning and traditional computer vision algorithms enables CogniPoint to run on low power consuming, lower cost ARM processors.
What is the pricing model for your technology and services?
There are two pricing components to our offering: the cost of the CogniPoint sensor hardware, and the data services generated by the sensor, for which we charge a recurrent monthly fee. This is aligned with the industry trend to move to a service-based model for end customers preferring OPEX expenses versus CAPEX. We offer the flexibility to balance the two components in order to accommodate client preferences.