Methodology for ADAS Validation: Potential Contribution of Other Scientific Fields Which Have Already Answered the Same Questions

Abstract

Since the 80s, the building of learn and test data bases for learning-based systems (i.e. neural networks) had to cope with problems of picking representative examples and measuring the generalization/the score of the system. And of course, real open world applications cannot be fully tested. It seems that artificial vision-based ADAS now discover the same question, and then, may use the same solutions, involving the same methodology (A.G.E.N.D.A.), using design of experiments and data analysis tools.