Automation never looked better

Seervision’s award-winning Adaptive Motion Control is built on machine learning, computer vision and model predictive control. Leveraging precise camera motion models, our software delivers perfect cinematography under all operating conditions. This enables new creative possibilities and high-quality footage for any production.

Header photo for our Automated camera software
Computer vision

Converting live feeds into intelligible objects

A fundamental part of Seervision’s technology, computer vision refers to the ability of our software to convert incoming frames of pixels into general semantics like “static person”, “moving person” or even identify ‘Bob’ himself in the frame. By processing sequences of frames in a batch, we are able to interpret scene dynamics as well as work through complex inference problems such as predicting a person’s intended movement path. The output of our computer vision module is the first input to all planning algorithms in the software stack.

  • Translating pixels into objects
  • Interpreting scene dynamics
  • Inference of human intentions
Model Predictive Control

Controlling any robotic system with mathematical models

Model predictive control allows us to create a unique, precise mathematical model for any robotic camera system in the industry. Based on these models we’re able to adapt all control algorithms to a particular system and take into account the acceleration capabilities of the motors, the optical characteristics of the lens and even the sensor size and weight of the camera body. What does it mean for you? Standardized performance no matter which PTZ camera you use or which camera body and lens combination you put on your robotic head.

  • Mathematical models for robots
  • Adaptive control algorithms
  • Robot and camera agnostic system
Machine Learning and AI

Customized implementation for each use case

Like most companies using AI and machine learning, we rely on data to train neural networks for our specific use cases. In the case of live video production, the goal is to tailor networks for object detection, object identification and semantic segmentation. When deployed for inference, these networks are responsible for providing the input needed by the path planning and framing optimization algorithms in our software pipeline. The system uses each production to learn how to make intelligent decisions regarding camera movements. Our users will soon be able to share specific video data with us and get a special version of the software, adjusted to their use case.

  • Networks trained on use cases
  • Unique identification of objects
  • Tailored to your own setup

Connect our back- and front-end with your own software

Communication between the Seervision Suite and the outside world happens in two places: the connection to the robotic system and the connection to the user interface. All our development is based on abstraction layers, splitting front-end and back-end to make sure all functionalities can be called by a third party system compatible with our application programming interface (API). We are happy to grant you access to the API to develop your own interface. We are also open to requests for extensions to the API and encourage all technically oriented clients and partners to push for the functionalities they need.

  • Distinctly split interfaces
  • Front-end and back-end protocols
  • Call functions with third party software
Web / Remote

Manage your whole fleet from anywhere in the world

It is common practice nowadays for intelligent systems to be connected with each other and be controllable from anywhere and anything. The whole Seervision Suite is developed with remote operation and web management in mind. Our control interface is web-based and makes it possible to command any robotic camera system from anywhere in the world directly from your phone, tablet or computer. Several functionalities of the Seervison Suite are already completely virtualized through docker which makes them deployable in custom servers or any cloud resource.

  • Fully connected robotic systems
  • Manage the camera fleet via the web
  • Deploy software on any server

Our approach to AI & automation

The term ‘artificial intelligence’ is nowadays overused, often taken out of context, and even feared. For us, AI is the term we use to explain the fact that our algorithms get better the more we use them and the more data we collect to train them. We believe AI and automation should take care of dull and repetitive tasks, allowing humans to focus on their creativity. We want to make visual storytelling easy and the solutions we create are aimed at simplifying complex tasks that stand in the way of that.