Shulou(Shulou.com)11/24 Report--

As a popular algorithm in the field of computer vision, 3D hand pose estimation and motion recognition are widely used in mobile robots and smart home scenes. Recently, CodaLab, the mainstream algorithmic science competition platform, held the "H2O: evaluation list of Interactive hand Motion recognition from the first Angle of View" Challenge. The Le Orange algorithm topped the rankings of "3D hand posture estimation" and "Motion recognition" with an accuracy of 0.0046 MJE (average joint error) and 97.11% accuracy, surpassing many first-class AI companies and academic research institutions.

(3D hand posture estimation)

(action recognition)

H2O (2 Hands and Objects) is the first hand-to-hand interactive multi-view RGB-D image sequence dataset proposed by the International computer Vision Conference ICCV2021. This list requires the entry algorithm to output the key points of the hand target in the video and identify the actions in the video by technical means. At the same time, we also need to adapt to the influence of environmental changes, camera perspective changes, motion blur, target occlusion, complex posture and other factors, and finally give the ranking according to the average joint error (MJE) of the left and right hands, the accuracy of action tags and other seven indicators.

As the world's leading intelligent Internet of things solution and service provider, Le Orange launched the self-developed AI algorithm matrix IMOU SENSE at the beginning of 2023, which is composed of intelligent algorithms, navigation algorithms and perceptual algorithms. In the evaluation, Le Orange uses two-stage method to optimize many links such as 2D attitude estimation and 3D attitude regression to improve the overall performance. Based on the framework of bone motion recognition, it optimizes many links such as bone map construction and spatio-temporal feature extraction to improve the overall performance. In the end, the seven indicators are ranked first, which once again confirms the cutting-edge level of Le Orange IMOU SENSE.

With the help of the motion recognition algorithm, the Le Orange camera can obtain the type of human action, and the device can deeply perceive the user's intention. In addition to gesture recognition, in the actual smart home scene, IMOU SENSE supports the accurate detection of people, cars, pets, faces, packages and other target categories, and supports any arrangement and combination of the above target categories. In addition, with the help of the target tracking algorithm, the camera equipped with IMOU SENSE can continuously track the target of interest, record the trajectory, and report the information in time when abnormal behavior is detected.

IMOU SENSE is used in Le Orange outdoor binoculars, indoor binoculars, indoor 4K and other new cameras. Based on the company's strong research and development capabilities, the recognition accuracy of Le Orange vehicle detection algorithm is as high as 99%, and accidents are fully captured without omission. At the same time, the probability of false positives is greatly reduced, the inconvenience caused by frequent alarms to users is reduced, and users can quickly obtain effective information.

At present, Le Orange algorithm has made major breakthroughs in many fields, such as target detection, target tracking, change detection, cutting-edge technology pre-research, etc., relying on IMOU SENSE's strong computing power foundation and leading algorithm capabilities, to create a more convenient and efficient experience for intelligent product users, so that care is more relaxed, security is more at ease.

In the future, Le Orange will continue to uphold the concept of "family safety expert", constantly polish its own technology, strive to improve its R & D capability, and bring safe and intelligent life to all parts of the world with the dual combination of software technology and hardware strength.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.