Can somone provide a working environment.yml for this notebook?

[BrainSlugs83]

RE: You do not need to install anything! You should be able to run all code from our released colab.
This was not my experience.

I've been at this for about a week now, and the Jupyter colab seems to requires dozens of packages that don't play nicely together, and I am not able to get very far with it due to tensorflow chicken & egg incompatibilties with all of the random packages that are required.

For anyone who sees this -- if this book is working in your environment -- can you just run a quick
conda env export --file environment.yml and share the file it produces?

Thanks!

[BrainSlugs83]

After banging my head against this for several more days, I finally found a combination of things that worked.

You can find an environment.yml on my personal fork of this repo, but it won't work as-is, because some software had to be installed with specific pip switches and git pathes that do not appear to be preserved in the environment.yml file. -- I have provided a full Windows setup guide on my personal fork of this repo.

I sincerely hope this helps anyone facing similar setup issues.

[ahron1]

Hi @BrainSlugs83 I'm trying to run this on Debian so I'm taking a slightly different approach from yours.

Superficially at least, it seems to me the function RunInferenceSingleImage definition is incomplete. All the later code expects a key detection_features in the output, which that function doesn't return. Maybe I'm missing something, but I didn't dig too deep on this yet.

I just raised a new issue about it.

Would you please share how you got that single image inference function working?

[BrainSlugs83]

Hey @ahron1, I'm not sure I ran into that issue myself. -- The issue I ran into was that the packages were not well specified anywhere, and I had to get really specific versions of everything (you can check your package versions against the ones in my yml file, and see if the package that provides that function has had a breaking change?)