HANGZHOU, China /September 27, 2018 As shown on The KITTI Vision Benchmark Suite official website, Dahua Technology’s SF flow estimation method took the 1st place and made two world records in KITTI Vision Benchmark Suite’s Scene Flow Evaluation 2015 and Optical Flow Evaluation 2015 competitions on September 25th, 2018 – with an SF-all and Fl-all of 7.58% and 5.45% respectively.
Dahua Technology’s SF Method Ranked #1 in KITTI Scene Flow Evaluation 2015
Dahua Technology’s SF Method Ranked #1 in KITTI Optical Flow Evaluation 2015
The KITTI vision benchmark suite is a systematic benchmarking platform designed to evaluate computer vision performance. Funded by the Karlsruhe Institute of Technology and the Toyota Technological Institute at Chicago, it is probably the world's first and largest benchmarking suite for vision based autonomous driving. KITTI includes real life images collected from a variety of scenery, from urban streets in the mid-size city of Karlsruhe to rural roads and highways. Each image contains sophisticated scenarios involving at most 15 vehicles and 30 pedestrians with varying levels of overlapping. The KITTI vision Benchmark suite comprises of real-world benchmarks for stereo, optical flow, visual odometry, object detection and tracking.
About Scene Flow and Optical Flow
The scene flow method can simultaneously estimate the 3D position and 3D motion vector of the object in the scene. It is the core algorithm for sensing the spatial geometry of the environment. When combined with the algorithm of sensing semantic information, it can understand the environment in a more comprehensive way. The estimation result of the 3D scene flow can be further decomposed into the 3D matching result and the optical flow matching result, and it can be considered that the optical flow is a projection of the scene flow to the two-dimensional image coordinate system.
Besides, the KITTI scene flow and optical flow competitions have provided Dahua Technology with an excellent opportunity to further its independent research and development of deep learning algorithms. Based on instance segmentation, depth parallax network and rigid body motion consistency constraint, Dahua Technology’s scene flow estimation algorithm surpasses other world-class AI companies and top academic research institutions, as well as scene flow research results in related ICCV, ECCV and CVPR papers. This symbolizes that Dahua Technology is in the lead in the fields of scene flow and optical flow algorithm.
The Application of the Winning Technologies
The technologies used in this competition have been applied to products such as dual-lens cameras and panoramic cameras, to improve the performance of target depth data capture and target separation of multi-lens cameras. Meanwhile, the technologies are also successfully applied to AR fusion applications, achieving AR semantic integration of single and multiple camera images, greatly improving the user experience of industry solutions.
Technological innovation is Dahua Technology’s driving force for development, and also its source of energy. Winning not one, but two KITTI competitions this time has for yet another time, showcased its world class technical research ability in the field of AI. With a mission of “Enabling a Safer Society and Smarter Living”, Dahua Technology will continue to focus on “Innovation, Quality, and Service”, to serve partners and customers around the world.