YOLO and CNN for Cat Detection and Recognition

This paper proposes a processing architecture for recognizing domestic felines in videos and images based on facial features. Some photos of the cats to be identified in the videos and images need to be collected in advance. A key aspect of this study is avoiding the need to retrain the model whenever the set of cats to be identified changes. The architecture involves five processing stages: video processing, YOLOv8 for cat detection, InceptionV3 for feature extraction, KNN for database searching, and voting process to improve classification. Transfer learning and fine-tuning techniques are employed to improve performance. The proposed method achieves high accuracy, offering a scalable and non-invasive solution for cat recognition, which can contribute to urban management, cat loss prevention, and medical insurance control for domestic cats, among other applications.