A deep learning-based approach for reconstructing 3D holographic images using autoencoder architectures. This project focuses on single-shot reconstruction to overcome twin image artifacts and reduce dependency on large training datasets, with applications in biomedical imaging and diagnostics.