Recently, the People's Government of Jiangsu Province issued the "Decision on Science and Technology Awards in Jiangsu Province for 2021" (Su Zheng Fa [2022] No. 28), which is the highest award in the field of science and technology in Jiangsu Province. According to the "Jiangsu Province Science and Technology Award Measures" and other regulations, a total of 263 projects were awarded in 2021, including 44 first prizes, 76 second prizes, and 143 third prizes.
Minivision Technology, in collaboration with Nanjing University of Technology, Harbin Institute of Technology, and Nanjing University of Posts and Telecommunications, jointly applied for the "Theory and Method of Image Restoration and Robust Recognition" and won the first prize in Jiangsu Province's Science and Technology. Yang Fan, CEO of Minivision, and Hu Jianguo, CTO, are the main implementers of this project.
The theory and methods of image restoration and robust recognition can solve industry challenges such as low quality, occlusion, and scarce training samples in AI scene based applications. Minivision Technology has achieved a series of challenges and challenges in image robust recognition, including effective restoration of low-quality images at the data level, robust recognition of occluded images at the representation level, and discriminant analysis in small sample situations at the feature level.
In response to the challenges of the data layer, Minivision adopts a combination of machine learning and deep learning strategies, such as multi-level continuous memory networks, deep adversarial generation networks, and other technologies; Techniques such as kernel norm matrix regression analysis and real-time data reconstruction were used to achieve robust image representation under occlusion; In response to the small sample training caused by difficulty in sample collection, a combination of deep learning methods such as OHEM and knowledge transfer, as well as machine learning methods such as heuristic feature fusion and integrated discriminant projection analysis, has greatly improved the robustness of feature representation in small sample training models.
At present, it is our province that is accelerating technological self-reliance and self-improvement, making greater efforts to build an independent and controllable modern industrial system, and striving to create a new stage of national important innovation highland, talent highland, and industrial highland. This award is a full recognition of Minivision Technology's research achievements in AI scenario based applications and artificial intelligence, and also a great encouragement for the company's independent research and development of algorithm algorithms and AI technology research work.
As a leading enterprise in the artificial intelligence industry in Jiangsu Province, Minivision will take innovation as the primary driving force for enterprise development, constantly making new breakthroughs in tackling key core technologies and accelerating the transformation of scientific and technological achievements. It will effectively shoulder the glorious mission of "striving to be a model, demonstration, and leading the way", and strive to write a new chapter of "strong wealth, beautiful heights" in the modernization construction of Jiangsu.
