F-ViTA leverages foundation models (SAM and Grounded DINO) to guide the visible-to-thermal image translation process using an InstructPix2Pix diffusion model. This approach improves translation accuracy and generalizes well to out-of-distribution scenarios.
