Great work! I'm a novice in the field of image recognition but do you think that using neural networks to represent both the text and image features in an h-dim

Great work! I'm a novice in the field of image recognition but do you think that using neural networks to represent both the text and image features in an h-dimensional embedding space could be extended to additional features like sound, or touch or images in other spectra like infra red? For example could we do better image recognition if there was also sound input from the scene or infra red in addition to visual spectra? It seems naively like lots of input data could be processed and then compared in this way. Thanks for taking the time to share your work and results.