Deploy Anywhere; Use Every Camera: The Power of the CrowdAI Platform
Patrick Collins and Taylor Maggos
In today's world, where we are surrounded by computers and cameras of all types and sizes, it's essential for machine learning services to be deployment-agnostic and camera-agnostic. Being able to work in any cloud, hardware, or software environment; and to use any camera or sensor is an invaluable advantage that has become increasingly important in recent years as the use of cameras has exploded in various industries. These features allow for greater flexibility and ease of use—exactly what CrowdAI strives to provide—enabling ML to be used in a wider range of applications.
The CrowdAI platform is an end-to-end solution for building domain-specific, customized computer vision models. On the platform, users can go from the ingestion of raw pixels to a production ready model, all in one place. CrowdAI built its platform to be deployable into any environment and to be used with media from any camera or sensor.
Being deployment-agnostic means that CrowdAI’s platform and models can be run in every major cloud and hardware environment; without requiring any specific services, bespoke work, or vendor lock-in. Users can choose the environments that best suit their needs and budget: everything from on-premises air-gapped servers to public and private clouds; enabling them to scale their applications as needed (Figure 1).
Figure 1: The CrowdAI platform can be deployed anywhere and everywhere: on the edge, through an API, to a private cloud environment and more. Our models have the same deployment agnostic capabilities— whether a customer wants to build the model on our cloud and deploy to the edge or do everything within their own environment, CrowdAI makes it possible.
CrowdAI’s platform is entirely managed with IaC (Infrastructure as Code), trivializing new deployments into even the most unusual of environments. CrowdAI’s full platform can be deployed within hours; all it needs is a cloud or a computer to connect to. We have achieved this notoriously elusive capability of deployment-agnosticity by first ensuring that the whole platform can run inside a generic Kubernetes cluster without relying on any cloud-specific features, storage backing, or services; and second by trivializing the creation of a kubernetes cluster in every cloud or environment. A deployment into AWS is nearly the same as a deployment onto bare-metal for us; with over 90% of the infrastructure code being identical. The remaining 10% is either DNS, storage backing, or the creation of the actual cluster; a task often simplified by tools like Rancher.
CrowdAI’s flexibility allows customers to achieve levels of privacy and capability unattainable with other machine learning platforms. When proprietary data is worth protecting; using a machine learning platform that can be deployed in environments entirely under your own control accelerates work on sensitive projects often limited by how few tools one’s data can be exposed to. Cost-savings also come easily; as CrowdAI’s models and platform can scale up and down to fit onto whatever hardware is available; whether that be edge computing devices, planes and drones with very considerable performance and power-draw limitations, or cloud environments where virtual machines are spun up only when needed. Using custom interfaces like gRPC, single containers deployed with one or more models receive media and deliver results using whatever mechanism is required or desired; further enabling deployment in limited environments that may come with serious limitations on how data can be input or output. Models can also be tuned to hit latency and frame rate targets even in these limited environments.
CrowdAI uses modern, industry-standard libraries and tools like TensorFlow, Docker, Kubernetes, and Terraform; enabling us to port our capabilities, models, and platform to new environments that may have not even existed when we designed our products to work everywhere.
Being camera-agnostic means that CrowdAI can work with a wide range of cameras or sensors, from low-resolution webcams to high-end professional cameras to satellites (Figure 2). This flexibility allows users to choose the cameras that best suit their needs and budgets; and to use cameras they already have instead of requiring new purchases.
Figure 2: Examples of cameras, sensors, and equipment CrowdAI works with to ingest media and build computer vision models (E.g. video surveillance cameras, unmanned drones, webcams, cell phones).
Wider compatibility enables a wider range of applications. Security systems often use a mix of different cameras within the same commercial or residential property; varying widely in terms of resolution, field of view, and other specifications. CrowdAI’s platform can build models that work with all of these different types of cameras, achieving a better result in less time.
The CrowdAI platform is also compatible with more complex cameras and sensors, such as those often used by governments and other industries in autonomous vehicles, drones, and robotics. Our camera-agnostic platform uses media captured by any camera—from a cell phone to a high-end satellite—allowing CrowdAI to build domain-specific, customized models for a wide range of uses and clientele (Figure 3).
Figure 3: Examples of models CrowdAI has built on a wide range of cameras and sensors used in various industries from commercial to government.
Agnostic on All Fronts
Machine learning is becoming an increasingly critical part of modern commercial and government processes, and as such, it is essential to ensure that these processes can function in a wide variety of environments. Ultimately, CrowdAI's platform allows for greater flexibility, versatility, and innovation in a wide range of industries by letting users develop and deploy their models anywhere and everywhere, using data from every camera and sensor. These capabilities make complex projects not just possible, but easy.
At CrowdAI, we’ve worked hard to build solutions that work for everyone, with everything, and from everywhere. While that’s a lot to say—and a whole lot more to actually do—we’re proud to walk the walk; getting far closer to those platitudes than seems possible not just for us, but for anyone. No matter how specific your requirements are or how limited your hardware is, whether your data is public on the internet or a matter of national security, whether you employ a team of experts or are a single individual who has never used or understood AI before, CrowdAI can make machine learning work for you.
Find out more about our platform and capabilities at CrowdAI.com.