Artificial vision intelligence innovator develops new software to simplify building and managing scalable vision intelligence solutions
Plainsight Technologies, the comprehensive vision AI factory, announced the availability of FilterBox. FilterBox is a secure runtime foundation that leverages MQTT, an IoT industry standard for data communication, that enables customers to translate raw video streams into a time series of business-relevant information to take action against. FilterBox is also the primary method of running Plainsight Vision Intelligence Filters which let companies of all sizes see more business as they measure and automate quality control, customer experience, logistics, inventory, and other critical business operations. The company also released two free demo Filters so anyone can easily try FilterBox.
Recommended: Coursera’s Generative AI Academy Enhances Executive, Foundational Literacy
“FilterBox makes it easy to build, deploy, and manage computer vision applications”
“FilterBox makes it easy to build, deploy, and manage computer vision applications,” said Logan Spears, CTO, Plainsight Technologies. “We’ve taken all the complexity out of building vision intelligence and wrapped it in a new concept called a filter, and aligned both our software architecture and our licensing model around that concept. This makes it super simple for customers to adopt and for our team to engineer scalable solutions for them.”
“Enterprises need an AI story, but this is easier imagined than implemented. In fact, in our latest research, we found thirty seven percent (37%) of organizations are adopting AI in their production cloud-based applications,” said Paul Nashawaty, Practice Leader and Principal Analyst, The Futurum Group. “It’s important that companies choose solutions that are future-proofed and that don’t lock them into a particular AI vendor ecosystem while it is still changing rapidly. Additional areas for users to seriously consider are the privacy of data – ensuring models aren’t trained from harvested corporate data — and supply chain security so they can enable innovation while managing risk.”
Recommended: CIO Influence Interview with Rosaria Silipo, VP of Data Science Evangelism at KNIME
Applying DevOps Best Practices to Video Intelligence at the Edge
FilterBox is based on Kubernetes, a cloud native computing paradigm which uses a low-resource variant called k3s that is efficient at the edge. Customers sensitive about data security, privacy, bandwidth limitations, cloud costs, or latency can run FilterBox collocated with the video streams. With FilterBox, customers can manage their own instance, maintain full control, and aren’t exposed to typical SaaS patterns of data being shared with a third party. FilterBox’s cloud native approach also allows best in class monitoring, logging, and observability with industry standard tooling.
Plainsight’s streamlined tools and services help organizations train, fine-tune, and optimize the accuracy of vision AI models without requiring machine learning expertise. Plainsight maintains a technology-agnostic approach by working with a variety of AI providers – including OpenAI, Google, Hugging Face, and more – helping to future-proof its customers in a rapidly changing AI landscape, and ensuring customers retain 100% ownership of proprietary video, image, and other data.
“FilterBox has the potential to open up new business opportunities for us,” said James Joslin, President and Principal Engineer at MarineSitu. “The ability to streamline our image processing is a huge win.”
FilterBox enables integrations into existing systems, assembly lines, programmable logic controllers (PLCs), or business intelligence tools organizations already use. Features in FilterBox 1.0 include:
- Containerized Runtime Engine for Filters – Allowing computer vision applications and AI models to be bundled and deployed using Docker, Kubernetes, Helm, and k3s (Kubernetes for Edge Computing).
- Pre-built Vision Intelligence Filter Repository – Choose from a library of pre-built vision intelligence filters designed to solve business problems related to data collection, alerting, and robotic automation of tasks. New filters are continually introduced.
- Supply Chain Security – FilterBox uses Chainguard Developer Images, which provides a software bill of materials (SBOM) for each container image and its open-source software components ensuring a trustworthiness across the software supply chain.
- Video Feed to Spreadsheet/Data Stream Pipeline – FilterBox connects to cameras or video streams over USB or network, and output usable data using industry-standard MQTT data. You can easily push this data to Snowflake, an AWS S3 bucket, Google BigQuery, or a database or data lake of your choice for further analytics.
- Private Filters in Customer Filter Repositories – Customers can access commercially licensed filters including variations for different AI models, hardware, cloud, or configuration.
Additionally, FilterBox is easily managed and controlled by IT in the cloud or on popular AI-enabled edge computing devices from providers like NVIDIA and SoftIron. And, the Command Line Interface (CLI) is developer-friendly, making it easy to automate deployment, configuration and updates using existing tools like Kubernetes, k3s, Docker, and Helm.
“As the world adopts AI workloads, they need a trusted platform to run close to where they collect data,” said Jason Van der Schyff, Chief Operating Officer at SoftIron. “Computer vision workloads dominate edge computing, and together with Plainsight Technologies, we can run FilterBox vision intelligence filters right inside HyperCloud to do anything from manufacturing quality control to drone-based mine surveying to scanning security footage.”
