Shulou(Shulou.com)11/24 Report--

Data tagging is ushering in a critical reshuffle moment. Recently, Quantum bit released the panoramic report on AIGC data tagging in China, which mentioned that the iterative process of large models from training to deployment of applications, the landing of many vertical scenarios of AIGC, and the exploration of cutting-edge areas such as general intelligence and embodied intelligence are all closely related to high-quality and professional scene data.

As a basic service, data annotation has never received as much attention as it is today. Especially as a new track synthetic data service, the potential market space is huge. According to the qubit think tank, the domestic market for AI basic data services will reach 10 billion yuan, accounting for about 10% of the global market. Among them, the synthetic data, as a new track derived from it, has a huge market space, with a growth rate of more than 40%, making it the fastest growing subdivision track.

The so-called synthetic data is generated by computer vision and graphics technology rather than through the collection of original data, which can be used to train, test and verify large models and algorithms needed by all kinds of agents. OpenAI CEO Sam Altman once said that all data in the future will become synthetic data.

As simulation data, composite data service has several significant advantages, the most important thing is to reduce cost and increase efficiency. According to the panoramic report of China AIGC data tagging, the representative of the typical service provider of synthetic data is the Coohom Cloud (Cluster Nuclear Cloud) platform of Group Nuclear Technology. As a synthetic data service platform, Coohom Cloud (Cluster Core Cloud) can improve the efficiency by 10 times. Based on the advantage of GPU cluster concurrent rendering, it can synthesize 200,000 sets of data per day, and based on the advantage of huge data sets, the training cost is greatly reduced.

In addition, composite data can supplement more edge and long-tail scene data, effectively solve the "data gap" in the large model era, and naturally avoid the problems of data privacy, security, compliance and so on. For example, it has related applications in the fields of data enhancement, model verification, interpretable AI, as well as autopilot, robot, biomedicine and so on. The qubit think tank predicts that the composite data will be the fastest growing track in the future, with an annual growth rate of 45%.

At present, as one of the few representative manufacturers providing indoor scene data services, Coohom Cloud (Group Nuclear Cloud) can synthesize 2D and 3D data sets for different application scenarios, such as intelligent robot, artificial intelligence, metasmology, intelligent real estate, self-driving, etc., with customers covering the world, serving many domestic and foreign technology giants such as Covos, Chase and Midea, and conducting in-depth cooperation with Intel on open-source projects such as industry and research.

One of the advantages of Coohom Cloud (Cluster Core Cloud) is that it has the largest indoor scene cognitive deep learning data set in the world, as well as Structured3D indoor environment cognitive data set, RPLAN household structure data set and so on. In terms of underlying technological innovation, the related technological research of group nuclear science and technology has been selected by SIGGRAPH, CVPR, ECCV and other international academic conferences for many times. In the industry vertical scenario, Coohom Cloud (Cluster Core Cloud) already has a leading edge.

Coohom Cloud (Cluster Core Cloud) already has two mature solutions. KoolAI is an AI training and synthesis data platform relying on the mass indoor 3D virtual scene of group nuclear technology, and relying on the cutting-edge laboratory of group nuclear technology, to provide enterprises with a variety of synthetic data of space scenes. SE is a cloud 3D indoor environment reconstruction platform for robots, especially indoor intelligent robots. Based on cloud computing and cloud service capabilities, it helps to redefine the cloud robot development and simulation process through realistic rendering, large-scale cloud computing clusters and cloud robot application deployment and management modules. In addition, the Coohom Cloud (Cluster Nuclear Cloud) team also provides enterprises with data customization and proprietary services to maximize the needs of various industries for data diversity.

At the beginning of this year, the SPEAR intelligent simulation platform developed by Group Nuclear Technology in conjunction with Intel Labs, the Spanish computer Vision Center and the Technical University of Munich was officially open to developers. This realistic rendering agent simulation platform can help developers speed up the training and verification of different intelligent robots, and promote the research and application of robots in human-computer interaction, digital twinning and so on.

As the application of artificial intelligence products becomes more diverse and complex, the industry has a stronger expectation for the flexibility and intelligence of AI services. How to provide high-quality data for AI applications in more industries and new scenarios is the next direction for Coohom Cloud (Cluster Core Cloud) to dig and explore.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.