OPPO received two awards for first place in the category AI in CVPR 2020, demonstrating technological capacity
Smartphone manufacturer OPPO took part in the conference on computational visual processing and pattern recognition (CVPR) 2020. OPPO demonstrated its advanced technology and innovative breakthroughs, especially in the field of AI. In addition, the company introduced a new Department of intelligent perception and interaction, which deals with the study of advanced AI technologies in its research Institute. Technological capacity OPPO did not go unnoticed. The company participated in three different competitions, winning two awards for first place and received two third places in the challenges CVPR 2020.
The technology that has won awards: the extremely high resolution, allowing you to make blurry picture 16 times clearer; visual localization, which can determine where the photograph by image analysis and recognition of human behavior on the video.
In addition, team members of research and development, OPPO has given lectures at several conference sessions and also participated in seminars where they could share their opinions about the technology of computer visual processing with thousands of leading researchers from around the world.
CVPR is the leading annual event for the achievements of the industry in computer generated visual processing, organized by IEEE (Institute of electrical and electronics engineers). The conference covers all aspects of computer visual processing and pattern recognition, machine learning and AI, and is of extreme value to students, scholars and researchers of the industry.
Eric Guo, chief research officer, Department of intellectual perception and interaction OPPO, said: “We are very pleased that this year in CVPR under the auspices of the OPPO we have achieved extraordinary results, demonstrating achievements in three main areas: the perception of the environment, computing, focused on human interests, and needs generation of visual learners. This is the first public demonstration of our R&D results since the establishment of the research Department of intelligent perception and interaction research Institute OPPO, which reflects the power of the AI technology of the brand”.
He added: “we All know that the invention of mouse and graphical user interface contributed to the success and popularity of the personal computer. Now touch screen deserves recognition for the success of smartphones. Whenever a human interaction with computer technology and intelligent perception is changing, this leads to completely new trends in technology. At OPPO we strive to make smart devices smarter, to better understand users. Then we develop an innovative method of interaction human-computer to start a new wave of technological progress.”
OPPO believe that cutting-edge research in the field of intelligent perception and interaction in the future will further expand the capabilities of the OPPO products and technologies that will provide users with a more innovative method of interaction and improved user experience.
Research centre OPPO the United States also participated in CVPR 2020 and took 6th place in the competition “Super-resolution of real-world track 1: Artifacts of image processing”. Meng zeebo, head of the laboratory Spacetime research center OPPO USA, gave a speech titled “Peering into the darkness from image to video”, telling about the study Institute regarding photography in low light using AI. Susue Quan, senior Director of the Department of computer visual processing and AI, made a presentation on localization and mapping, entitled “creating a unified system of SLAM, which runs on the smartphone and AR glass from scratch.”
Challenge “extreme Perceptual superresolution”
In this competition, and OPPO research team had to solve the problem: to achieve extreme super-resolution for single image with magnification x16. The goal is to obtain models that can create high-resolution images with best quality — one that best reflects reality. The team developed and demonstrated the effectiveness of neural network called RFB-ERSGAN surpassed 280 other participants, suggesting a more intelligent analysis of effective evaluation and more diversified methods. In the end the team won first place, ahead of team Facebook, Baidu, Microsoft, Amazon and Chinese Academy of Sciences.
Neural network RFB-ERSGAN OPPO creates a more natural texture, for example, animal fur, building the structure and texture of plants
Technology will be used for a number of scenarios. For example, it can help users to edit LR photos taken by their smartphones on the HR image with incredible detail. Other programs can include repairing old or damaged photos and old videos or a detailed analysis of satellite images and space images.
Challenge “Visual localization for handheld devices”
OPPO is developing AI programs for image analysis in different dimensions
Taking first place at the external visual localization (Aachen Day-Night) and third place at the visual localization in space (InLoc), OPPO has created a pipeline of monocular visual localization, which uses semantic and root hints to find the exact location of any image by request.
It is expected that the future use shall include the application of augmented reality, intelligent robotics and accurate navigation. For example, when the AI-assistant equipped with a camera, it can use the technology’s visual localization for the analysis of images and achieving precise positioning to move to the right place.
Detection of activity in the extended video
The aim of the program detect activity in a dynamic video where the OPPO team ranked third, was to develop algorithms for automatic detection of activity. OPPO has invented their own algorithms for the efficient identification and reorganization of the characters in each frame and recognition of dozens of human actions in complex dynamic videos. This technology will play an important role in the area of interaction human-computer and sports video analysis in the future. For example, the future smartphone will be able to automatically detect moving objects, and accordingly to adjust the focus of the camera.
OPPO develop algorithms for the efficient identification and reorganization of human behavior in complex dynamic video
A new Department of intelligent perception and interaction
In addition to successful performances at the conference CVPR 2020, research Institute of OPPO announced that earlier this year created a new Department of intelligent perception and interaction, marking a new opening, the company OPPO in the field of advanced AI technology. The Department attracts talent from many of the leading technology companies and research institutes. For the position of chief scientist in the field of intelligent perception also invited a leading expert on AI Eric Guo, who conducted a workshop on how to identify drones and tracking materials at the conference CVPR.
Eric First worked at Microsoft as a researcher and for many years conducted several industry-leading research. His model of computer visualization is widely used in many industry solutions.
OPPO has created a team to study AI in 2015 with a focus on the areas of phonetics and semantics, and computer visualization. As of March 2020, OPPO have filed more than 1,900 patent applications in the field of AI throughout the world. It will support OPPO in the era of 5G and even 6G by the development of completely new devices and provide a reliable technical basis for the future era of intelligent connectivity.
Today OPPO has gathered a research team of more than
10 000 people, has established six research institutes and four research centers around the world. In addition, the founder and CEO of OPPO Tony Chen said the company in the next 3 years will invest 50 billion yuan (7 billion US dollars) expenditure on R&D for development of technology hardware, software, and 5G, AI, AR, big data and other related technologies. OPPO constantly seek to explore advanced technologies to provide customers with next generation innovative methods of interaction and user experience.