Machine Learning for Object Detection and Classification in Deep-Space Astronomical Images

CNNs, Vision Transformers (ViTs), and Capsule Networks (CapsNets) are used in this study to enhance astronomical image classification for accurate object identification. Pre-trained models like ResNet50 and EfficientNet optimize feature extraction, while pre-processing, transfer learning, and synthetic data augmentation increase flexibility. In deep-space images, the hybrid model that combines both approaches performs better than either one alone, increasing robustness and classification accuracy. According to a comparison analysis, this approach reduces feature loss and misclassification, making it perfect for largescale astronomical surveys.