InnoPeak Technology
InnoPeak Technology, based in Palo Alto, California, conducts advanced research in smartphone technologies, focusing on computer vision, video, and image processing to enhance user interactions with smart devices, communication networks, and cloud services.
Company Overview
InnoPeak Technology specializes in cutting-edge research in smartphone technologies, focusing on computer vision, video, and image processing. Based in Palo Alto, California, the company strives to enhance end users' daily interactions with technology through advanced algorithms and products. They work with smart devices, communication networks, and cloud services.
Research and Development
InnoPeak Technology is known for its innovative approaches in various research areas. These include real-time lighting estimation for augmented reality, multi-person 3D pose estimation, continuous-touch text entry for AR glasses, and more. The company has developed a method for real-time lighting estimation from a single image using deep neural networks and differentiable screen-space rendering.
Technological Contributions
InnoPeak Technology has introduced multiple groundbreaking technologies. They proposed PoP-Net for predicting multi-person 3D poses from a depth image, achieving state-of-the-art results. The company designed Continuous-touch T9 and Continuous-touch Dual Ring text entry interfaces for AR glasses and created a pipeline for reconstructing 3D human avatars from a single image using GAN-based texture inference.
Innovative AR Solutions
InnoPeak has combined geometric information from Visual-Inertial Odometry (VIO) with semantic information from object detectors to enhance mobile AR experiences. They also developed RGBD-based globally-consistent dense 3D reconstruction with online texturing and introduced methods for occlusion handling and collision detection in smartphone AR using ToF cameras.
Advanced Image and Video Processing
The company has developed GIA-Net for low-light imaging, integrating global information to improve performance. InnoPeak introduced GCF-Net to boost video action classifiers with minimal computation overhead, and RCA-GAN for image super-resolution and noise reduction, delivering better visual quality and performance.