Project:	Evaluate Generative AI in Artistic Illustration of Space Innovative Advanced Concepts (JPL AO No. 15313)
Disciplines:	Artificial Intelligence, Art, Illustration
Mentor:	Virgil Adumitroaie, (JPL), Virgil.Adumitroaie@jpl.nasa.gov, Phone: (818) 393-7038
Background:	Visual media has a significant value when communicating concepts at all stages of a project, from the proposal phase to the outreach, when discoveries are shared with the public. It is often said, "A picture is worth a thousand words.” But how would a scientist, technologist, or engineer produce that valuable illustration if one lacks artistic talent, even if one knows the exact 1000 descriptive words? And how do we know which words our pictures really communicate to others? Answers to these questions are becoming possible with the recent Generative Artificial Intelligence (generative AI) model that produces unique photorealistic images from text and image prompts. [1] “Text-to-image generative AI products such as DALL-E 2, Stable Diffusion [2], and Midjourney now regularly conjure images based on text descriptions and are now capable of generating photorealistic images of exceptionally high quality” [3]. The reverse is also possible: for example, Image to Text AI asticaVision provides an extremely detailed description for any image [4]. Also, Salesforce - Blip Image Captioning is a cutting-edge AI model that bridges the gap between images and text by generating accurate and meaningful captions from images [5],[6].
Description:	A selection of space advance concepts will be made from a list of NIAC, and Blue Sky concepts, as well as results from simulation models using data obtained by recent JPL spacecraft such as Juno and Ingenuity. Will create alternative illustrations using DALL-E and Stable Diffusion, starting 1) from descriptors made by the involved scientists and 2) based on descriptors indicated by Chat GPT from abstracts of papers. The resulting illustrations will also be evaluated by two methods: 1) descriptors by people not involved in the studies (summer interns) and 2) by Salesforce BLIP Image captioning to generate descriptors. An analysis will be made on the correlation of the resulting descriptors with the original descriptors used in the text-to-image step. An iterative technique will be explored that will result in the generation of images that, when interpreted, will result in almost identical descriptors as used in image generation.
References:	[1] https://jaredzimmerman.medium.com/narrative-style-creation-transfer-with-llms-text-to-image-generative-ai-systems-646a79901e5b [2] https://aws.amazon.com/what-is/stable-diffusion/ [3] https://www.practicalecommerce.com/6-image-to-text-tools-ai-powered [4] https://spectrum.ieee.org/ai-art-generator-smarter-inference [5] https://github.com/salesforce/BLIP [6] https://www.plugger.ai/models/salesforce-blip-image-captioning
Student Requirements:	Professional background in art, with a focus on illustration. Demonstrated exposure to space concepts. Prior experience with Generative AI is desired but not required.
Location / Safety:	Project building and/or room locations: . Student will need special safety training: .
Programs:	This AO can be done under the following programs:   Program    Available To        SURF@JPL    both Caltech and non-Caltech students  Click on a program name for program info and application requirements.