Nokia Facilitates Machine Vision Deployment via Bring Your Own Algorithms and Federated Learning

Subscribe To Download This Insight

1Q 2021 | IN-6076

In the face of a crowded competitive landscape, Nokia is trying to differentiate itself through innovative features, such as Bring Your Own Algorithms (BYOA) and federated learning. This ABI Insight looks at these features in depth and evaluates Nokia’s value propositions

Registered users can unlock up to five pieces of premium content each month.

Log in or register to unlock this Insight.


Nokia's Foray into Machine Learning -Based Machine Vision


In 2017, Nokia acquired SpaceTime Insight, which included applications for asset-intensive industries and an integrated platform for developing, running, and creating applications. Now, as part of Nokia Cloud and Network Services, Nokia SpaceTime analytics provides an extensive portfolio of advanced analytics capabilities.

An example of this is Nokia’s video analytics solution called SpaceTime scene analytics, which allows users to perform surveillance, anomaly detection, and metadata generation based on sensor inputs, images, and video streams. The solution supports a wide range of video cameras and IoT sensors through local-area and wide-area networks. In order to ensure the support for legacy systems, Nokia’s SpaceTime scene analytics can be integrated with all major Vendor-Management Systems (VMS) such as Milestone, Genetec, and Eagle Eye Networks, as well as legacy SCADA and ERP systems. While these are very common features, Nokia SpaceTime has some other innovative features up its sleeve.

Innovative Features


One key innovative feature of Nokia’s SpaceTime scene analytics is Bring Your Own Algorithm (BYOA). Enterprise users can develop their own Machine Learning (ML)-based machine vision application in a software container and host it on gateways or call it from public cloud using the Application Programming Interface (API) available in the scene analytics Software Development Kit (SDK). The scene analytics SDK supports common ML algorithm development tools such as Jupyter Notebook and TensorFlow. This provides a high degree of flexibility and customization for enterprises. However, it would be interesting to see how Nokia addresses the need of ML developers when it comes to specific chipset architecture support, such as Nvidia’s Graphics Processing Unit (GPU), Intel’s Movidius Myriad X Vision Processing Unit (VPU), and Xilinx Versal ACAP.

Another feature is federated learning. Initially popularized by Google, federated learning is an ML technique that enables training of a DL model across multiple edge devices without the need for sending training data to a centralized model in the cloud. Instead, the shared model in the cloud aggregates the learning from all edge devices. The key benefit of federated learning is data privacy. Only minimal data transfer will take place because each camera will collect its own data for self-learning and share only the new learning instead of the entire data input with the shared model in the cloud. At the same time, the shared model in the cloud will be updated with the learning from all the cameras and become significantly accurate over time, without any compromise to the cybersecurity of monitoring or security system.

More important, SpaceTime allows Nokia to offer an analytics solution on top of its end-to-end connectivity solutions. In recent years, Nokia has been actively pushing for the adoption of private LTE and 5G networks in enterprises. Offering an analytics function to the private wireless network is a great value add. By leveraging edge computing architecture, the analytics layer can be hosted at on-premise servers or micro data centers, allowing the deployment of “dumb” cameras and keeping the video traffic local. In many enterprise sectors, such as transport and logistics, manufacturing, port management, and smart city and spaces, there is a strong focus on end-to-end integration and data security and privacy. Coupled with Nokia’s end-to-end portfolio of private LTE and 5G network solutions, Nokia is able to offer enterprises support from not just the Nokia network deployment team, but also its data science and professional services teams. This minimizes the touchpoints that the client needs to have during the implementation phase and accelerates the time-to-market.

Challenging Competitive Landscape


There is little doubt that Nokia is facing severe challenges in this space. Not only is it competing against video analytics specialists and highly rated video analytics startups such as SenseTime, Clarifai, and Paravision, the company is also up against cloud service providers that have end-to-end portfolios in cloud video analytics solutions. Increasingly, the cloud service providers are also moving into the edge computing domain, such as AWS’s Snowcone and Snowball, where both sides have many overlapping and similar functions.

Nonetheless, Nokia does offer a unique value proposition. Nokia has been able to demonstrate its capability to implement multiple applications, including telecommunications, public safety, traffic management, and environment monitoring on a single platform and visualize the information on a single pane of glass. Since SpaceTime scene analytics also supports federated learning, the solution is capable of ensuring no video data will be streaming to a centralized server because all information will only be stored locally for model training. This addresses cybersecurity concerns in industries where end users are very cautious with centralized processing and storage.

Another way Nokia can potentially expand its market influence and brand recognition is to join industrial alliances, such as Open Security and Safety Alliance (OSSA). OSSA aims to define shared measures and specify reference designs for data security and privacy in the security and surveillance market. This is done through developing a single common Linux-based platform for all products and creating an open marketplace for all vendors. By joining alliances like this, Nokia can play a role in influencing the formulation of open standards and hardware specifications to support their software performance, while helping their clients to reduce fragmentations and frictions during deployment through common platform.