An image information fusion based simple diffusion network leveraging the segment anything model for guided attention on thermal images producing colorized pedestrian masks

Suranjan Goswami, Satish Kumar Singh

Abstract: Highlights•Thermal Infrared and RGB image fusion method using diffusion (DDPM) effective for pedestrian detection.•Tight pedestrian masks generated via guided segment anything model (SAM) can be used as ground truth for learning the segmentations.•End to end deep learning pipeline for guided attention on pedestrians.•Mean squared error (MSE) is stable as a generative error term for guided attention for fusion taskss.