On November 18, Shanghai Artificial Intelligence Laboratory, together with Shangtang technology sensetime, the Chinese University of Hong Kong and Shanghai Jiaotong University, jointly released a new generation of general vision technology system "Scholar", which aims to systematically solve a series of bottleneck problems in the current artificial intelligence vision field, such as task commonality, scene generalization and data efficiency. ▲ source: Shanghai Artificial Intelligence Laboratory At present, the technical report "intern: a new learning paradigm towards general vision" has been released on arXiv platform. The general vision open source platform opengvlab based on "Scholar" will also be officially open source at the beginning of next year to disclose the pre training model and its use paradigm, data system and Evaluation Benchmark to academia and industry. According to relevant technical reports, a "Scholar" based model can fully cover the four visual core tasks of classification, target detection, semantic segmentation and depth estimation. Shanghai Artificial Intelligence Laboratory said that compared with the current strongest open source model (clip released by openai in 2021), "Scholar" has made a significant improvement in accuracy and data use efficiency. Specifically, based on the same downstream scene data, the average error rate of "Scholar" on the 26 data sets of the four tasks of classification, target detection, semantic segmentation and depth estimation has been reduced by 40.2%, 47.3%, 34.8% and 9.4% respectively. It home learned that the general vision technology system "Scholar" (Intern) is composed of seven modules, including three infrastructure modules: general vision data system, general vision network structure and general vision evaluation benchmark, as well as four training stage modules to distinguish upstream and downstream. (Xinhua News Agency)