That is going to make it very difficult to manage. The tif files are big as it is, and you could write them all out to csv, but it is going to be clumsy and large.

To answer your question, just read the tif files as e.g. in the starter. Then if you want you can convert the data to e.g. pandas dataframe and write out or write out using numpy write to text functionality. But note earlier comment.

I think you need to rather consider and plan a bit on what you want to do, then that will perhaps make it more clear how to proceed. I think you have the right pieces of the puzzle, but you are putting them together wrong or you are missing a few pieces. Perhaps you want to load the tif, do some processing and then write to csv for example.

Yo can generate a unique csv file where each row represents a pixel value from a tiff file. Then you can aggregate pixels per field_id by computing their mean or random sampling some of the pixel values of an specific field id as a regularization technique. Generating a one-time csv file to play with will allow you fast experimentantion.