Reliable engineering
takes many forms

Automated Visual Inspection comparing Computer Vision to Machine Learning

11 Novembro 2019by admin
by J. Gomes-Mota¹; Albatroz Engenharia¹
CIRED2019, Madrid.

Competing solutions for automated classification of visual issues are coming from two different realms: computer vision and machine learning. The author relates on the experience of his team with methods from both camps to detect and classify visual Points of Interest as per request of transmission and distribution utilities. The main novelty is the comparison of the performance of different methods applied to the same data sets and terms of reference. At this stage, no side has yet won but some relevant conclusions are already drawn: long term asset management tools are required to meet long life spans, humans keep key roles in the learning loop; modularity goes hand in hand with long term efficiency.