Improved U2Net-Based Surface Defect Detection Method for Blister Tablets

• Published in: Algorithms Vol. 17; no. 10; p. 429
• Main Authors: Zhou, Jianmin, Huang, Jian, Liu, Jikang, Liu, Jingbo
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.10.2024
• Subjects: Accuracy; Algorithms; blister tablets; Blistering; Classification; Deep learning; defect detection; Defects; Drug dosages; Drugs; Feature extraction; Gaussian Laplace operator; Good Manufacturing Practice; Image processing; Image segmentation; Laplace transforms; large kernel attention mechanism; Methods; Pharmaceutical industry; Product quality; Quality management; Semantics; Surface defects; Surfaces; Surfaces (Technology); Tablets; U2Net; China

Aiming at the problem that the surface defects of blAister tablets are difficult to detect correctly, this paper proposes a detection method based on the improved U2Net. First, the features extracted from the RSU module of U2Net are enhanced and adjusted using the large kernel attention mechanism, s...