Combination Of CNN And LSTM In Cat Breed Detection In Pet Stores

Abstract

The research examines the utilization of the Long Short-Term Memory (LSTM) algorithm for identifying cat breeds in pet retail establishments. With numerous cat breeds globally, each with unique needs and behaviors, manually identifying them presents a challenge. This research aims to address this issue by implementing deep learning techniques for breed classification, which will assist pet owners and storekeepers in recognizing and caring for cats more efficiently. The dataset used consists of 9,755 images, spanning six distinct cat breeds: Siamese, Birman, Russian Blue, British Shorthair, Bengal, and Egyptian Mau. The methodology encompasses image preprocessing, partitioning the dataset into training, validation, and testing subsets, and training the LSTM model for cat breed classification. The model is included into an Android application created with the React Native framework, utilizing Python for backend model processing. This study's results are anticipated to yield a reliable and precise instrument for cat breed identification, with possible uses in pet store administration and individual pet care. The ultimate application is assessed according to accuracy, precision, recall, and F1-score, all of which are obtained from the confusion matrix.