I had a large geotagged UAV orthophoto. Since the Roboflow platform did not support the tif format, I converted the tif orthophoto to jpg, split the large orthophoto into many small tiles, and performed annotation in Roboflow for training, testing, and validation dataset. I applied Yolo and predicted the results with bounding boxes as well. Now I need to proceed further in my work however, I need to geotag the predicted image and predicted bounding boxes as to match its location in the original tif orthophoto. However, in using the RoboFlow platform, I only used the images that contained better datasets, so I selected the images based on the quality of datasets in the images for annotation. Now I am not able to convert it to the geotif images to take it to the original location as I have many images and I don’t have the origin or the top left corner pixel of each tile. Any alternative solutions are highly appreciated. Please help me solve this problem.