The input image is supplemented with depth and semantic descriptions derived from the image using foundation models. The proposed RAIDS model has a fully convolutional architecture that incorporates a ...