Jiangsu Province issued a work plan to speed up the construction of embodied intelligent robot datasets.

Recently, the Jiangsu Provincial Department of Industry and Information Technology issued the "Accelerating the Construction of Embodied Intelligent Robot Data Sets Work Plan", focusing on broadening the collection scene to solidify the source of data supply, strengthening innovation support to enhance the quality and effectiveness of data empowerment, improving the standard system to regulate data circulation and transactions, and deepening collaborative linkage to build a sound development ecology. It proposed 11 specific measures to accelerate the seizing of strategic opportunities for the development of artificial intelligence terminals. High-quality data sets are the key foundation for the training, reasoning, and validation of large-scale artificial intelligence models. They are data sets collected, cleaned, classified, and labeled through intelligent processing according to specific standards, with corresponding mechanisms for updating and maintenance. This work plan detailed the division of responsibilities for various departments and put forward specific work requirements and goals. For example, it aims to guide the opening of real application scenarios, with no less than 100 low-sensitive, high-value real application scenarios opened annually; support embodied intelligent robot whole-machine enterprises and downstream users to form alliances, enter industrial manufacturing, livelihood services, special environment and other fields to carry out data collection in real scenarios, with a maximum support amount of 1 million yuan for eligible individual projects; accelerate the integration test platform construction in the field of embodied intelligent robots, including data adaptation, algorithm iteration, whole-machine testing, and scene verification, and encourage leading enterprises and academic institutions to jointly carry out collaborative innovation of "standardized data sets + lightweight industry models + intelligent entities"; and focus on key processes such as data collection, preprocessing, annotation, and quality evaluation, pushing forward the revision of no less than 30 related standards in a classified and stratified manner.