Automated Classification System for Tick-Bite Defect on Leather

Table 3

Classification performance comparison to the state-of-the-arts on the same dataset.

Method	Preprocessing	Feature Extractor	Classifier	Accuracy (%)
[6]	Grayscale normalization, Gaussian filter	Statistical features (mean, variance, skewness, kurtosis, lower, and upper quartile value)	SVM	74
[23]	Grayscale normalization, resizing, Canny edge detection, block partition	ANN		80
[23]	Grayscale normalization, resizing, Canny edge detection, block partition	AlexNet		76
[4]	—	Mask R-CNN		80
Ours	Histogram matching, resizing, grayscale normalization, Gaussian blurring, and Canny edge detection	Histogram of gradient	k-NN, SVC, SVM, MLP, AdaBoost, decision tree, random forest, discriminant analysis, extreme gradient boosting	94

The bold numbers represent the highest values within the experimental results presented.