Employees:
101~200
Year Established:
2021
Business Type:
Manufacturer
The company was established in June 2021. It is a high-tech artificial intelligence enterprise focusing on the research and development of 3D visual software and hardware products Industry. The company is committed to applying the world's leading 3D geometric deep learning technology to the field of industrial intelligence, enabling machines to complete complex environments based on 3D.
The positioning, recognition, detection, guidance and other tasks of 3D vision can be used to empower intelligent manufacturing with a new generation of 3D vision.
The company's R&D personnel account for more than 70%, and the core team is led by world-class authoritative experts in the field of 3D artificial intelligence. Members have doctoral/master backgrounds in artificial intelligence from top universities at home and abroad.
The hardware and productization team comes from Philips, ASML, Tencent, Honeywell and other well-known technology companies, with rich experience in the development and implementation of intelligent manufacturing products, and strong technical strength
Participated in the IROS Object Pose Estimation Competition and won multiple championships and runners-up.
Teams including SenseTime and Tsinghua University developed 3D detection models such as F-ConvNet and HotSpotNet, and maintained the No. 1.
The plan to overwhelm Huawei and other corporate organizations.
Developed the DualPoseNet full-degree-of-freedom attitude estimation model, whose accuracy ranks first in the NOCS benchmark database, significantly ahead of other solutions.
Developed the SRDC domain adaptation learning model, and maintained the first accuracy on the unmanned driving transfer learning benchmark database from synthetic data such as GTA5->Cityscapes and SYNTHIA->Cityscapes to real scenes
Participated in the IEEE BTAS 2016 Video Person Recognition Evaluation Challenge and ranked first.
Participated in the IEEE BTAS 2016 Video Person Recognition Evaluation Challenge and ranked first.
Published GPNet and other models to achieve deep learning non-registration object grasping, performance surpassed the model developed by Nvidia, and published the largest simulation non-registration grasping data set
The world's first 3D AffordanceNet 3D function affordability analysis method and large-scale benchmark data set to help academic and industrial human-computer interaction research and development
Researched and developed SkeletonNet, ToMoNet and other networks, for the first time realized deep learning complex topological surface generation, and was selected as the best paper candidate of CVPR 19, the top conference of artificial intelligence
Initiated Analytic Marching non-destructive analytical surface mesh theory and algorithm, and open source AnalyticMesh software, which is expected to replace Marching Cubes and become a new industry standard for grid extraction and storage
Propose a deep model adaptive optimization method, break through the barriers between traditional MVS technology and deep surface reconstruction, and greatly improve the effect of MVS reconstruction
