Concept: American startup eCapture has launched an AI-powered 3D stereo depth camera, LifeSense G53. A VGA quality, small form factor camera uses the advanced vision technology developed by eYs3D, a Taiwan-based end-to-end computer vision platform provider. Its tiny size and lightweight profile aids in depth acquisition and object tracking for robotics, industrial, and other AI-driven applications.

Nature of Disruption: The stereo vision devices like LifeSense G53 embed the eYs3D vision processing SoC (system on a chip) to compute stereo depth data, without having to rely on the GPU/CPU at the host. The G53 provides a 50-degree field of view (FOV) and includes two mono sensor pairs for various resolutions of stereo, mono, and depth disparity/distance map output via USB. At the heart of the camera is a cutting-edge high-performance image processing chip that allows it to capture the same frame on both sensors, providing data for the development of visual SLAM algorithms and reducing mistakes caused by unsynchronized frames. This fast processor enables the device to create a full stereo depth map without relying on the host’s CPU/GPU. As a result, 3D depth-sensing applications will benefit from a more integrated and low-power solution. It employs Global Shutter, a technology to realize simultaneous exposure of all pixels of both cameras. In rapid-moving shooting scenarios, the technology can effectively reduce image distortion to provide a more accurate image. In addition, it has two built-in active IR illuminators, allowing for visual binocular plus structured light fusion. This enhances recognition accuracy for white walls and untextured objects and also enhances vision in completely dark environments.

Outlook: In light of recent uncertainty within the Intel RealSense product line, eCapture expects to release a full range of depth map cameras soon to address the growing need for stereo imaging technology and to take market share. It also has plans to release G100 in Sept’21. Moreover, the startup offers software development tools for its products that allow for ease of integration into Windows, Linux, and Android OS environments. The camera has an open API and is designed to work with MediaPipe (Google) to efficiently train new vision recognition models. The product can be used in applications such as smart retail, facial recognition, object detection, iris tracking, hand tracking, 3D scene learning, robotics, drones, and autonomous guided vehicles. The startup claims that several leading consumer robotics companies recently incorporated eCapture cameras with depth vision to their cleaning robot design.

This article was originally published in Verdict.co.uk