Greg Gandenberger

I have solved the immediate problem described above by changing line 391 of auxil.py from `filename = tkFileDialog.Open(filetypes=filetypes,title=title).show()` to `filename = tkFileDialog.Open(title=title).show()` (removing the `filetypes` argument). I take it that...

That seems right. For the moment I am just going to live with this solution; I do not know when or if I will get the chance to develop something...

Megadetector has separate "animal" and "human" categories, so we could use it out of the box to classify images as empty, containing humans, or containing animals.

Go for it! I did a first test using Megadetector to classify images as containing humans or not and as containing animals or not; those results look good. I am...

Right now I believe our main user is comfortable with consuming JSON dictionaries of probability scores, but I'm open to adding a web interface for other potential users.

Nothing is provided for writing to a file. I think that's fine -- whatever tool a user is using to hit the endpoint (e.g. bash, Python, R) will have its...

Great! I would love to move to GitHub Actions. The Dockerfile is in https://github.com/uptake/autofocus/blob/master/autofocus/predict/Dockerfile. It is specific to the app that serves predictions from a trained model.

Seems like a reasonable thing to try! I am not very familiar with SSIM, so I am not sure how suitable that particular metric is for this problem. I am...

The percentage of "empty" images is not low -- somewhere around 30%, IIRC.

FWIW, [Autofocus](https://github.com/uptake/autofocus) checks for conformity to both `flake8` and `black`. Rather than using a pre-commit hook, we ask contributors to ensure conformity to both tools as part of the PR...