Material Classification with a Transfer Learning based Deep Model on an imbalanced Dataset using an epochal Deming-Cycle-Methodology

This work demonstrates that a transfer learning-based deep learning model can perform unambiguous classification based on microscopic images of material surfaces with high degree accuracy. A learning-enhanced was successfully used in combination an innovative approach for eliminating noisy data automatic selection using pixel sum values, which refined over different epochs to develop and evaluate effective classifying microscopy images. The evaluated achieved 91.54% accuracy the dataset set new standards method applied. In addition, care taken achieve balance between robustness respect model. Based this scientific report, means identifying could evolve support identification, suggesting potential application domain materials science engineering.