Phages are most likely to be the most abundant biological entity on the planet. Phage particles are simple in that all known phages are comprised of 2 components: 1) genetic material; and 2) structural proteins. Despite their simplicity and abundance, up to 56% (151784/269906) of phage coding sequences have unknown function based on the annotations of coding sequences retrieved by GenBank queries. As a result, the use of sequence similarity is not an effective method of detecting phage structural proteins from large amounts of unknown phage sequence data that is generated by metagenomics. In addition, phage structural protein function is challenging to predict from sequence data because phage structural proteins possess little known conserved catalytic sites or sequence domains. To detect phage structural protein sequences we used multiple two layer feed-forward Artificial Neural Networks (ANN) that use the Levenberg-Marquardt supervised learning algorithm with validation. The trained ANNs correctly identified viral structural protein genes with >87% sensitivity and >80% specificity.