I would like to do something quite simple but it seems that it’s not simple at all.
I have a set of images and the corresponding ground truth mask for semantic segmentation. To make it more clear I attach a couple of images. As you can in the left column the two images depict exactly the same scene, top one is the RGB image while bottom one is the ground truth where the different class instances are coloured accordingly (black is for soil, green for crop and red for weeds). I would like to extract the orthomosaic of the area from the RGB images and have an overlay of the ground truth, as in the right column images.
I assume this is a simple process by texturing the orthomosaic grid (calculated from the RGB input) with the corresponding images from the ground truth data. Yet I have no idea how can this be done. I thought it would be possible to provide the two different modalities (RGB/ground truth mask) as bands of a multispectral image, yet ODM do not recognize them as multi-camera setup. As far as I understand XMP tags have to be defined, I spent a couple of days and still I have not managed to add somehow the corresponding XMP tags to activate the multi-camera functionality.
Do you have any idea/tips on how to texture the extracted grid or how to add correctly the required XMP tags? Any thoughts are more than welcome.