Image Classification of Occluded and Non-Occluded Cats and Dogs Datasets with Machine Learning

Pet classification performance can be significantly affected by occlusions, such as pillows, hands, blankets, and shadows. Existing research on pet recognition and classification focuses on poses, breeds, and natural scene variations, but there has been limited work on occlusion-based pet classification. This project aims to improve the performance of deep neural networks for pet classification by using data augmentation, transfer learning, and fine-tuning techniques. Two independent datasets of occluded and non-occluded images of cats and dogs are used to evaluate the proposed algorithms. The investigation explores the impact of transfer learning on cat and dog classification using two proposed models: a fine-tuned Convolutional Neural Network (CNN) and a pre-trained VGG16 model. By leveraging advanced techniques such as transfer learning and fine-tuning, this research demonstrates the potential for significantly improving pet image classification performance, especially in scenarios involving occluded images. These findings contribute to the broader field of computer vision and provide valuable insights for further research in pet image analysis and classification.