Pengsi technology takes aiot ecological platform as the productivity core of integrating emerging technologies and enlarging the efficiency of information infrastructure. In terms of industrial integration, Pengsi technology takes the construction of a new human centered smart city as the starting point, and relies on Pengsi aiot cloud platform to build a full video analysis system, portrait big data system, smart community system, intelligent traffic system, etc., deeply empowers urban public security and governance, intelligent human settlement scene and other fields, and creates a new infrastructure in the inclusive era.
In addition, Shen shengmei, chief scientist of Pengsi science and technology and President of Singapore Research Institute, was invited to attend the summit and delivered a keynote speech entitled development and application of visual intelligence based on transfer learning, which elaborated the pain points of traditional machine learning, the breakthrough brought by deep learning, deep transfer learning, and some technologies and examples used in transfer learning The demand of transfer learning in visual intelligence is analyzed.
Shen shengmei first pointed out that one of the pain points of traditional machine learning is poor generalization ability, which leads to poor robustness, high landing cost and poor user satisfaction. The breakthrough of deep learning model for computer vision is also trained on the given data set, which can well reflect the characteristics of given data. Although the model trained in these specific datasets has high accuracy, far beyond the traditional machine learning, its performance can not be maintained in new scenarios and new tasks. Therefore, the breakthrough of deep learning still depends on the data.
In addition, Shen shengmei also pointed out three major pain points in the implementation of deep learning AI
The perception environment and application scene are changeable: weather, light, angle, occlusion and other factors lead to different imaging quality and poor quality; inconsistent training data and landing scene lead to the performance of AI model in vain;
The retraining model needs a large number of AI professionals to complete, with long cycle, high cost and slow effect, which has become an obstacle to AI inclusive;
AI talent shortage and high cost will hinder enterprises to adopt AI and its benefits.
Shen shengmei said that Pengsis mission is AI is service, which can solve the above pain points and realize inclusive AI. Pengsi realizes Ai Service Online + AI function customized terminal + self closing ring ecology through the ecological technology deployment of aiot platform. General intelligence is the inevitable trend of the development of the next generation of AI. Pengsi will take root in the actual situations of transfer learning, unsupervised learning, self supervised learning, small sample learning and multimodal learning, which truly reflect the landing of AI.
The goal of Pengsi technology is to walk in the forefront of international computer vision in terms of algorithm technology. Based on the development of cutting-edge technology based on AI landing, build cloud AI online migration learning platform, strengthen the lightweight design system of end-to-side AI, and promote tob / TOC product innovation oriented by user value.