<![CDATA[Cat breed classification is challenging in image processing due to complex visual variations from crossbreeding, which affect care requirements. This study evaluates the effectiveness of Contrast Limited Adaptive Histogram Equalization (CLAHE) and Histogram Equalization (HE) in cat breed classification using a VGG16-based Convolutional Neural Network (CNN). The dataset consists of 4,656 cat images from six breeds, processed with CLAHE and HE for contrast enhancement before training. It is divided into 70% for training, 15% for validation, and 15% for testing. The model is trained for 10 epochs using the Adam optimizer, a 0.0001 learning rate, and batch sizes of 16, 32, and 64. Evaluation using accuracy, precision, recall, and F1-score shows that CLAHE achieves the highest accuracy (99.39%), surpassing HE (99.17%) by 3.29%. CLAHE is more effective in preserving local details, improving precision (78.67%), recall (78.33%), and F1-score (78%). The highest performance is in the Sphinx breed (F1-score 92%), while the lowest is in American Shorthair (F1-score 72%). A high standard deviation indicates classification variations across breeds, but CLAHE consistently improves model accuracy. These findings suggest that CLAHE is more effective than HE in enhancing cat breed classification and offers a more efficient solution than adopting a complex model architecture.]]>
Copyrights © 2025
