This document presents a study on developing deep neural encoding models to predict fMRI responses in the human visual cortex to natural visual scenes. The study uses a large dataset of fMRI responses from 8 subjects viewing 73,000 natural images. Various pre-trained deep neural networks are evaluated as feature extractors to map images to visual features. A two-step voxel-based encoding approach is proposed, applying dimensionality reduction before training linear regression models for each voxel. The best models achieve high prediction accuracy across visual cortical regions of interest, demonstrating the effectiveness of transfer learning from computer vision models for the image-fMRI encoding task.