Astro-VAE Pipeline Enhances Faint Object Detection in Space Imagery

A deep-learning pipeline uses Tiny-U-Net for stellar masking and a partial-convolution variational autoencoder for context-aware background inpainting. This approach suppresses structured sky backgrounds and dense stellar fields to improve low signal-to-noise ratio object detection in optical space situational awareness imagery.