Por Rodrigo Santos (orientadores: António Branco / João Silva).

The merging of language and image processing has supported increasingly successful multi-modal solutions thanks to recent advances combining the strengths of both fields. Editing an image on the basis solely of natural language instructions is a particularly enticing challenge at the forefront of this endeavour.

In this seminar, I will summarise my previous work on Instruction-guided Image Editing, whereby image-tokens are manipulated through Large Language Models. Additionally, I will discuss my ongoing research on Unsupervised Instruction-Guided Image Editing with Diffusion models.