We build computer-vision software that has to work outside the lab, on aircraft, spacecraft, ground vehicles, and embedded hardware, where the camera is the primary sensor and no operator is standing by to reset it. We wrote the camera-calibration tutorial that ranks first on Google, and helped put visual navigation into GPS-denied environments where a wrong pose estimate crashes a vehicle far from human rescue.
Computer Vision and Vision Navigation run from camera calibration and multi-camera geometry through visual and visual-inertial odometry (VO/VIO), SLAM, and sensor fusion across cameras, IMU/INS, LIDAR, structured light, and depth sensors. We help build SAR processing and analysis tools, structure-from-motion and point-cloud reconstruction (including Gaussian splats and NERFs), and detection and tracking using machine learning where it beats classical methods and plain OpenCV where it doesn't.
Some of the problems we've helped solve: vision-based navigation for an autonomous lunar lander with no GPS and no prior map; on-orbit vision navigation for space domain awareness; a Kinect-class depth-sensor toolkit; SAR analysis tooling. If your camera has to make a decision, we have likely solved a version of it.
We run OpenCV and CUDA, write in C++ and Python, target real-time and embedded hardware, and wire perception into OpenGL, OpenSceneGraph, Vulkan and VulkanSceneGraph when the result has to be seen and not just computed, including the VR, AR, and xR systems where perception has to stay locked to the world.
Who works on it
AlphaPixel is a US-owned small business, founded in 2004, with seven senior developers who each have 25-plus years in 3D and imaging. DLA DD2345 / ITAR registered. We take closed-source defense work and open source alike, and we do not sell a product, so we have no reason to point you at one.
Frequently asked questions
What kinds of computer vision problems does AlphaPixel work on?
Vision that has to work outside the lab, on aircraft, spacecraft, ground vehicles, and embedded hardware where the camera is the primary sensor and no operator is standing by. That runs from camera calibration and multi-camera geometry through visual and visual-inertial odometry, SLAM, and sensor fusion, plus detection and tracking, SAR analysis, and 3D reconstruction. If your camera has to make a decision, we have likely solved a version of it.
Do you use OpenCV, or machine learning and deep learning?
Both, chosen by which one wins. We use machine learning for detection and tracking where it beats classical methods, and plain OpenCV where it does not. We are not selling a model or a framework, so the method we pick is the one that fits your accuracy, latency, and hardware budget.
Can computer vision run in real time on embedded hardware?
Yes. We write in C++ and Python, run OpenCV and CUDA, and target real-time and embedded platforms, not just workstations. When the result also has to be seen, we wire perception into OpenGL, OpenSceneGraph, Vulkan, and VulkanSceneGraph.
Do you do visual odometry, SLAM, and sensor fusion?
Yes. We build visual odometry and visual-inertial odometry (VO/VIO), SLAM, and fusion across cameras, IMU and INS, LIDAR, structured light, and depth sensors. This is the same toolchain behind our GPS-denied navigation and space domain work.
Can you build 3D reconstruction like photogrammetry, Gaussian splats, or NeRFs?
Yes. We do structure-from-motion and point-cloud reconstruction, including Gaussian splats and NeRFs, and we have built depth-sensor toolkits for Kinect-class hardware. We can hand you a mesh, a point cloud, or a live reconstruction wired into a real-time scene.
Can you work on defense or export-controlled computer vision under NDA?
Yes. AlphaPixel is DLA DD2345 and ITAR registered and takes closed-source defense work and open source alike. Much of our vision work has reached the field indirectly through primes and government programs.
How is hiring AlphaPixel different from a general AI or ML shop?
We are seven senior developers, each with 25-plus years in 3D and imaging, and we do not sell a product, so we have no reason to point you at one. You get people who have already put vision on real vehicles in the field, not a demo that only works on a clean dataset.
How do we engage AlphaPixel for a computer vision project?
As a subcontractor, a team-augmentation partner, or the hired specialists for the one hard perception piece. Describe the problem through our contact page; scope and cost depend on the work, so we quote after we understand it.
Talk to us
Tell us what you’re building and where it’s stuck. We’ll tell you straight whether it’s something we can help with, and how we’d approach it.
Or see how we run consulting and software development engagements.
