Precise Image Classification with Xception Model
Bhoomika Bhoomika · 2024
Aiming to improve accuracy and dependability in image classification tasks, this work investigates the deployment of a finely tuned Xception model for identifying cat and dog photos. Utilizing a dataset of 14,000 photos (7,000 dogs and 7,000 cats) obtained from Kaggle and other reliable sources, the study used data augmentation methods to enhance model performance and variability. With respect to precision and recall values of 1.00 and 0.91 for cats and 0.90 and 1.00 for dogs respectively, the finely trained Xception model attained an amazing total accuracy of 0.95. With 0.95 for both classes, the Fl-scores show balanced performance over categories. Effective learning without overfitting was suggested by the study of training and validation losses; moreover, the trends in training and validation accuracy revealed constant improvement. With low misclassifications and great dependability, the confusion matrix confirmed even more the strength of the model. This work emphasizes the Xception model's performance in pet classification as well as its possible uses in automated pet identification and allied domains.