| Abstract [eng] |
In lumber industry most of the visual quality inspections are still done by trained operators, this is tedious and repetitive task with high likelihood of human error. Currently, new automated solutions with high resolution cameras and visual inspection algorithms are being tested, but they are not always fast and accurate enough. This paper describes how highly effective faster region-based convolutional neural network (Faster R-CNN) is implemented in wood veneer defect detection. This method was successfully used in medicine and surveillance solutions, but there are no papers about this method usage in wood veneer quality assurance. In this paper, research on method accuracy dependence from learning epoch count, learning and testing proposal region count is carried out. Performance optimization using different base models and proposal region count is also analyzed. Experiments using synthetically augmented dataset and additional transfer learning classifier are also performed. |
