T/SAITA 001-2021 Artificial intelligence and computer vision system evaluation specification

Scope

Main technical content: The "Evaluation Specification for Computer Vision Systems Based on Machine Learning" stipulates the evaluation standards for computer vision systems based on machine learning, and is applicable to the design, R&D, testing, deployment, and operation and maintenance of computer vision systems based on machine learning. Other fields of artificial intelligence can also refer to and use it. The standard sorts out the evaluation indicators of computer vision systems from the aspects of functional effectiveness, performance, compatibility, maintainability, portability, training data quality, adversarial samples, robustness, interpretability, and security, and proposes an evaluation indicator system and evaluation process specification for computer vision systems. In particular, the standard addresses the security issues of artificial intelligence by adding indicators with artificial intelligence characteristics such as the impact of adversarial samples, interpretability, robustness, and security to the traditional software quality indicator system, providing evaluation guidance for the security evaluation of artificial intelligence.

Development Information

Related Standards