Open kadyb opened 2 years ago
It may be better (i.e. easier) to drop the spatial reference at the model training stage and save the files as .png (instead of geotiff). Alternatively, the spatial reference can be saved in separate files as worldfiles (.wld) using GDAL.
https://www.tensorflow.org/api_docs/python/tf/io/read_file https://www.tensorflow.org/api_docs/python/tf/io/decode_png https://gdal.org/drivers/raster/png.html
https://www.tensorflow.org/api_docs/python/tf/keras/utils/image_dataset_from_directory
It may be better (i.e. easier) to drop the spatial reference at the model training stage and save the files as .png (instead of geotiff). Alternatively, the spatial reference can be saved in separate files as worldfiles (.wld) using GDAL.
https://www.tensorflow.org/api_docs/python/tf/io/read_file https://www.tensorflow.org/api_docs/python/tf/io/decode_png https://gdal.org/drivers/raster/png.html
https://www.tensorflow.org/api_docs/python/tf/keras/utils/image_dataset_from_directory