April 4, 2018 The artificial intelligence "national team" cloud made a major breakthrough in technology from the cross-mirror tracking technology (ReID). At the same time, in the Market-1501, CUHK03, DukeMTMC-reID three data sets from the University of Technology, Nanyang Polytechnic, Institute of Automation, Chinese Academy of Sciences, Tsinghua University and other well-known universities, enterprises and research institutions stand out, refreshing the world record.
The highest hit rate (Rank-1 Accuracy) on the Market-1501 reached 96.6%, breaking the previous world record created by Ali iDST in January 2018, enabling cross-mirror tracking technology (ReID) to reach the first time in accuracy. At the commercial level, artificial intelligence is about to cross from “brushing face†to a new era of “recognizing peopleâ€.
The first hit rate is 96.6%. Cloud has renewed three world records from the technology of cross-mirror tracking (ReID) technology.
Leading algorithm refreshes three records simultaneously
Person Re-IdenTIficaTIon (ReID) is a popular direction of computer vision research. It is a technology that uses computer vision technology to determine whether a specific pedestrian exists in an image or video sequence.
Simply put, it can “recognize†you without looking at the face, just by wearing, posture, and hairstyle.
This technology can be used as an important supplement to the face recognition technology. It can continuously track the cross-camera for pedestrians who cannot obtain clear-cut faces, and enhance the spatio-temporal continuity of data. This technology can be widely used in video surveillance, intelligent security, intelligent business and other fields.
Market-1501, CUHK03, and DukeMTMC-reID are the most authoritative data sets currently measuring ReID technology. The first hit rate (Rank-1 Accuracy) and Mean Average Precision (mAP) are the core indicators for measuring the level of ReID technology.
The first hit rate is 96.6%. Cloud has renewed three world records from the technology of cross-mirror tracking (ReID) technology.
The first hit rate is 96.6%. Cloud has renewed three world records from the technology of cross-mirror tracking (ReID) technology.
Cloud Technology has set the best level in the industry in all three data sets. The Rank-1 Accuracy in the Market-1501 dataset has reached a staggering 95.7%, and the Re-Ranking technology has reached 96.6%. mAP is a more comprehensive measure of the effect of the ReID algorithm. It requires that the searched image and the retrieved image match correctly, not just the first hit. Cloud Technology has raised the best level of Market-1501's mAP index by nearly 5% to 86.9%, and reached 94.2% after using Re-Ranking technology. It is able to achieve such a large breakthrough, fully demonstrating the value of the research results of Cloud from Science and Technology ReID, which will inevitably promote the substantial progress of ReID technology, and also make ReID accelerate the practical application.
Why do we need cross-target tracking (ReID) technology?
Face recognition technology has matured over the past few years and has been applied in many scenes and products. However, face recognition technology only uses human facial information, but does not use other information of the human body, such as clothing. , posture, behavior, etc., in addition to the application must be able to capture the face, which can not be satisfied in many scenes, such as bow, back, blurred body, hat occlusion and so on.
Cross-mirror tracking (ReID) technology can make up for the lack of face recognition technology. Cross-mirror tracking (ReID) technology can recognize pedestrians based on pedestrian wear, posture, hairstyle and other information. This raises the cognitive level of artificial intelligence to a new stage, and now cross-mirror tracking (ReID) has become an important research direction in the field of artificial intelligence.
However, the existing research results are not very mature, and there is still a certain distance from the actual commercial requirements. And Cloud has made a major breakthrough in the technology of cross-mirror tracking (ReID) technology, and will raise the existing technology level to a new stage, which will greatly promote the progress of technical research and application landing in the industry, and will also greatly promote artificial intelligence. “Brushing the face†has entered a new era of comprehensive “living peopleâ€.
What are the difficulties in cross-mirror tracking (ReID) technology?
Cross-mirror tracking (ReID) technology is similar to face recognition technology. There are many difficult points to overcome, such as light, occlusion, and image blur. In addition, pedestrians wear a variety of clothes, the same person wears different clothes, different people wear similar clothes, etc., also put forward higher requirements for the cross-frame tracking (ReID) technology.
The alignment technique that is widely used by pedestrians to cause widespread use on the human face also fails in cross-mirror tracking (ReID). Pedestrian data acquisition difficulty is much more difficult than face recognition data acquisition, and pedestrian information is much more complicated than human face. These two factors are superimposed to make the algorithm research of cross-frame tracking (ReID) more difficult. It is also more important. Breakthroughs in the effective design of algorithms and the reduction of data dependencies to achieve cross-mirror tracking (ReID) are the consensus in the industry.
This cloud provides a very good idea for solving the ReID problem by proposing the integration of pedestrian global information and discerning multi-granularity local information. Cloud from the technology proposed this program has several advantages (1) compact structure: the program achieves end-to-end direct learning, and does not add additional training processes; (2) multi-granularity: the integration of pedestrians' overall information and Multi-granularity details of the degree of discrimination; (3) attention to detail: the model truly understands what is human, the model will focus on the knee, clothing trademarks, etc. can significantly distinguish some of the core information of pedestrians.
“Brushing face†is an important application in the field of computer vision, and “studenting people†will prompt the computer vision industry to enter a new stage of development. Cloud has been deeply researched in many subdivisions of technology in the direction of “peopleâ€, such as pedestrian detection, pedestrian structured information extraction, human key point detection, attitude estimation, and behavioral motion recognition. The landing of technology will enable everyone to realize the help and improvement of intelligent security, human-computer interaction, automatic driving, intelligent business, and home life.
Absolute rotary Encoder measure actual position by generating unique digital codes or bits (instead of pulses) that represent the encoder`s actual position. Single turn absolute encoders output codes that are repeated every full revolution and do not output data to indicate how many revolutions have been made. Multi-turn absolute encoders output a unique code for each shaft position through every rotation, up to 4096 revolutions. Unlike incremental encoders, absolute encoders will retain correct position even if power fails without homing at startup.
Absolute Encoder,Through Hollow Encoder,Absolute Encoder 13 Bit,14 Bit Optical Rotary Encoder
Jilin Lander Intelligent Technology Co., Ltd , https://www.landerintelligent.com