Rockchip RV1108 visual processor with CEVA-XM4 DSP handles Evomotion’s high-resolution real-time stitching on-board at 70% lower power consumption compared to CPU/GPU implementations.
CEVA, Inc., the leading licensor of signal processing IP for smarter, connected devices, and Evomotion Technology, Ltd, one of the leading deep learning and computer vision combining hardware solution providers in the industry, announced that the companies have collaborated to enable advanced imaging and computer vision features for Evomotion’s ROD-1/360°/720° dual-lens spherical camera. The ROD-1 is capable of taking 3k resolution 30fps videos and 4k resolution photos with panoramic stitching in real-time, implemented on the CEVA-XM4 imaging and vision DSP, eliminating the need for post-processing on a smartphone or PC. The camera is now available and is one of the best value 360 degreecameras in the market today.
The camera utilizes Rockchip RV1108 visual processors, which include the CEVA-XM4 DSP to handle advanced imaging and computer vision tasks. Evomotion ported a range of imaging algorithms and high-performance stitching to the fully-programmable CEVA-XM4 using CEVA’s development environment and tools, supported by CEVA and Rockchip engineers. Following this successful project, CEVA and Evomotion intend to extend the partnership to include broader imaging solutions for dashcam, sports camera, SLAM and real-time 3D reconstruction in the future.
Lufang Gao, CEO of Evomotion, commented: “With the increasing demand of edge processing and embedded intelligence products, a powerful and power efficient smart vision platform being able to handle computer vision and deep learning algorithms is needed for our customers and the industry. The CEVA-XM4 DSP and Rockchip’s RV1108 platform deliver best-in-class performance for us to continually deliver customer value to the market.”
Weibing Huang, CTO of Evomotion, added: “We were extremely impressed with the quality and capabilities CEVA’s software development environment, allowing us to easily and efficiently develop our algorithms. In addition to the stitching, the CEVA-XM4 also handles distortion, color correction and other image enhancement features, with notably low power consumption. While achieving the same stitching performance, the power consumption of CEVA-XM4 is 70% lower compared to a CPU/GPU implementation.”
Ilan Yona, vice president and general manager of the Vision Business Unit at CEVA, commented: “Evomotion’s 360 degree camera is an excellent endorsement of the strong computer vision capabilities of our CEVA-XM4 DSP, seamlessly handling multiple algorithms and creating outstanding video quality from dual camera streams stitched together in real-time. Our goal is to bring computer vision and deep learning capabilities to every camera-enabled mass market device, and working directly with innovative companies like Evomotion who deploy our technology is an essential step to achieving this.”
CEVA’s latest generation imaging and vision DSP platforms address the extreme processing requirements and low power constraints of the most sophisticated machine learning and machine vision applications used in smartphones, surveillance, augmented reality, Sense and Avoid drones and self-driving cars. These DSP-based platforms include a hybrid architecture composed of scalar and vector DSP processors coupled with a comprehensive Application Development Kit (ADK) to streamline software deployment. The CEVA ADK includes: CEVA-Link for seamless software level integration with a host processor; a range of widely used and optimized software algorithms; the CEVA Deep Neural Network (CDNN2) real-time Neural Network software framework which streamlines machine learning deployment at a fraction of the power consumption of the leading GPU-based systems, and; state-of-the-art development and debugging tools. For more information, visit http://www.ceva-dsp.com/app/imaging-computer-vision/.
Evomotion Technology, Ltd.