- About Us
Dahua’s PSPO Method Ranked #1 in KITTI Flow 2015 Benchmark
KITTI vision benchmark suite is a benchmark to evaluate computer vision performance. Established by Karlsruhe Institute of Technology and Toyota Technological Institute at Chicago, it is the world's first and largest benchmark for vision based autonomous driving. KITTI includes real images collected from a variety of road scenes, from urban streets to country roads to highways. Each image contains a sophisticated scenario involving, for instance, a crowded vehicle and pedestrians, with various levels of occlusion. It comprises of real-world benchmarks for stereo, optical flow, scene flow, visual odometry/SLAM, 3D object detection, 3D tracking and road/lane detection.
Dahua’s Central Research Institute, based upon stereo &multi-frame technologies, utilizes planar models together with deep information, and established the multi-degree scene model demonstrating pixels, segments, planes and targets through finer labeling and analysis of the static background in the scenes. This is how Dahua’s PSPO 3D Scene Flow Method is developed. Dahua PSPO method just refurbished the world record, beating fellow participants in KITTI Flow 2015 Benchmark include Max Planck Institute for intelligent systems (MPI)、Swiss Federal institute in Zurich (ETH Zurich)、TU Graz institute of computer graphics and vision (ICG), as well as well-known university labs from the University of Tokyo, University of Toronto and Brown University. This symbolizes a breakthrough in Dahua’s 3D scene analysis &movement anticipation technology based upon mobile platforms, laying a solid foundation for the development of products such as automatic driving and intelligent moving robots.
About 3D Scene Flow
3D scene flow algorithms are able to simultaneously estimate the three dimensional location and motion vector of objects within a scene. It is the core algorithm used to determine spatial geometric structures when mobile platforms perceive their environment, enabling various applications in location, navigation, and accident aversion. It can also provide other forms of assistance within the semantics of environmental perception. The estimations offered by 3D scene flow algorithms are able to be mapped across 2D image spaces, engaging in perceptive observation by means of 3D matching and optical matching.