"Pioneering Trials + Exemplary Cases" – Our City's High-Quality Datasets Selected for National Lists

Recently, the National Data Administration released the list of pioneering trials for high-quality dataset development and the list of exemplary cases of high-quality datasets. Our city's Geely Automobile Research Institute's "Geely Automotive Xingrui High-Quality Dataset" and Bluetron’s "Multimodal High-Quality Industrial Dataset for the Green Petrochemical Industry" were included in the pioneering trials list, accounting for 50% of the provincial total. The Konfoong Biotechnology's "High-Quality Cervical Cell Dataset" was selected for the exemplary cases list, making it the only one in the province.

Leveraging the Geely Xingrui Intelligent Computing Center as its foundation, the Geely Automobile Research Institute plans to construct a high-quality dataset exceeding 100 petabytes in scale. This dataset will encompass seven categories of data: supply chain data, ecological product data, telematics signal data, artificial intelligence voice data, multimodal unstructured data, intelligent driving data, and vehicle simulation data. It will be used to support the development and training of the Geely Xingrui large model and Geely's intelligent driving models, thereby enhancing vehicle intelligence, user experience, and supply chain/ecological product collaboration capabilities.

Relying on the supOS operating system, the Bluetron has collaborated with ecosystem partners to efficiently integrate multidimensional data from the green petrochemical industry, including production data, experimental data, and public data. The company plans to build a multimodal high-quality dataset comprising over 5TB of text, 5TB of video, over 100,000 hours of audio, and over 500TB of video data. This dataset will be used to develop industrial large models, focusing on application scenarios such as process R&D and design verification, intelligent product quality inspection, intelligent maintenance and management of industrial equipment, production safety control, supply chain collaboration, and operational excellence. These efforts aim to help enterprises improve production efficiency, accelerate technological innovation, and promote industrial transformation and upgrading.

Addressing issues such as low efficiency in cervical cancer screening and insufficient service capacity of primary medical institutions, the Konfoong Biotechnology has broken through the limitations of traditional manual testing by digitizing pathological slides. The company has built a cervical cell image dataset of approximately 200TB and developed an AI-assisted screening model for cervical cancer based on artificial intelligence. This model enables automatic localization, identification, and statistics of suspicious cell areas, significantly increasing screening volume and accuracy. Based on this achievement, the company's self-developed cervical cancer auxiliary diagnostic system "Cervical Cell Scanning and Analysis System" was successfully selected as domestically produced first-set equipment.

In recent years, our city has consistently treated data as a key element for developing the digital economy and a core element for advancing artificial intelligence. We have deeply researched pathways for constructing high-quality datasets, actively selected and promoted benchmark demonstrations for high-quality dataset development, and studied and introduced specialized policy clauses to facilitate the construction of high-quality datasets in various industries. As a result, a number of high-quality datasets with industry representativeness have been initially formed. For the next step, our city will accelerate the promotion of national pioneering trial projects for high-quality dataset construction, diligently implement the provincial action requirements for high-quality dataset development, create a series of iconic industry-specific high-quality datasets, and strive to explore and form the "Ningbo Experience" in high-quality dataset construction.