Greg.Randall

February 21st 2025

I shoot film sometimes, though haven’t managed to do it in a couple of years. First, we were on septic, and I wasn’t going to be putting development chems into the yard. Second, I’m a bit nervous about having the chemistry around my son.

In any case I’ve had an idea kicking around for a few years that you might be able to use machine learning to help figure out film and developer dilutions, times, and temperatures. The Massive Dev Chart mostly has you covered, but occasionally I’d run into an odd combination or do a push/pull that wasn’t on the chart and so I’d need to refer to the push compensation. Occasionally I’d do a push compensation and then also crank up the heat to keep the times from being interminable. I always made an image but was it the best I could have gotten? I mostly didn’t test enough to find out, though I did test one enough to make a submission to The Massive Dev Chart, which they emailed me back to make sure I hadn’t made a typo in the temperature.

I haven’t gotten to the machine learning part of the project, but I wrote some code to collect data. I’ve got some cleaning up of the data going on to do things like drop all entries that don’t have a full complement of information and averaging numbers where it makes sense like when it says 200-300 ASA/ISO, that gets averaged to 250. I’ve also dropped all of the two stage developer entries, which isn’t ideal but it’s only about 200 out of 16,000. If you want the full-full set of data review the file ‘all-film-all-developer.csv’, though do note it’s full of duplicates.