Add bodyweight exercises

[b3nj5m1n]

This adds a bunch of bodyweight exercises, enough to cover most of the main progressions of the reddit recommended routine.

I did my best to research the targeted muscle groups but obviously can't promise anything, I used an LLM for writing most of the instructions to avoid copyright issues and touched them up a bit myself, but they're still far from perfect. I think most people are probably not using this as their only resource for learning new exercises though, so hopefully that's fine.

Because I couldn't find your code for generating the exercise images, I've left those blank for now.

In some cases, I think we should probably expand some of the available options in app/src/main/java/org/librefit/enums, for example by adding a pull-up bar or rings to the equipment, although existing exercises are also affected by this (Pull-ups currently require no equipment, for example). The other thing would be hip flexors in the muscle groups, that's something I kept encountering while researching and couldn't find an appropriate option for currently.

One area for improvement here would be encoding the information for progressions trees, but maybe that's outside the scope of this project.

Thank you very much for all your work on this project!

[IamDg]

Hi @b3nj5m1n,

Thank you for taking the time to contribute!

Unfortunately, I haven't defined exactly how to contribute to the dataset (I'll fix it soon). However the reddit source looks great!

Regarding the images, they were generated locally using ComfyUI. If you are comfortable and able to generate them, I can share exactly how they were made. :)

I agree that some equipment is missing, like the pull-up bar (this is the most important problem IMO).

Finally, the progression tree would be a very nice addition (I'd be the first to want it) but it would be hell to implement properly. :(

Now, some guidelines to contribute to the dataset:

• No duplicates (and of course no ID duplicates)
• Must follow schema
• There should be at least one image that is described accurately (at least in part) by the instructions

Let me know what you think!

[b3nj5m1n]

Regarding the images, they were generated locally using ComfyUI. If you are comfortable and able to generate them, I can share exactly how they were made. :)

Unfortunately, ComfyUI doesn't seem to be available on Linux, so I'm afraid I won't be of any help there.

I agree that some equipment is missing, like the pull-up bar (this is the most important problem IMO).

I've now added pull-up bar & rings as equipment options, specifically in the Equipment enum, json schema, translation strings & formatter, hopefully those are all the necessary places.

Finally, the progression tree would be a very nice addition (I'd be the first to want it) but it would be hell to implement properly. :(

I totally understand.

---

I've written up descriptions for the two exercises where I had previously just put YouTube links, I still think the YouTube videos would be much more useful though. Do you not want any links in general, or would it be fine to add them back in addition to the written instructions?

[IamDg]

Unfortunately, ComfyUI doesn't seem to be available on Linux, so I'm afraid I won't be of any help there.

Are you sure? I could be wrong, but maybe it's this: https://docs.comfy.org/installation/desktop/linux. However, what I meant for this is to generate images without copyright, so cloud models could be a solution.

I've now added pull-up bar & rings as equipment options, specifically in the Equipment enum, json schema, translation strings & formatter, hopefully those are all the necessary places.
They look great! Also, the progressions as different exercises are a pretty good workaround.

I've written up descriptions for the two exercises where I had previously just put YouTube links, I still think the YouTube videos would be much more useful though. Do you not want any links in general, or would it be fine to add them back in addition to the written instructions?

I'm not actually against it. My only concern is how to actually display them (they should already display as clickable links) and which videos/channels are chosen as the source. The ideal would be some sort of channel with a consistent design, format, length, and so on (and if I remember correctly, there are YouTube channels like that).

Overall, you did a great job! The last step would be adding the images as indicated in the new guidelines for submissions in the dataset.

[b3nj5m1n]

Are you sure?

There's no distributed version available, I found something for my distro which is marked as probably broken, I'm trying it anyways, but it's been compiling for a few days now, so I'm slowly loosing hope. (Edit: it just failed after about 20 hours of compile time)

I've started generating some images using cloud models, but because of rate limiting this might take a while.

In the meantime, I've added a couple more exercises, so I think this now covers all the main progressions of the RR, though some of the alternative progressions are still missing.

[IamDg]

Happy to hear this!

But yeah I agree: cloud models offer the best quality; just be sure their generated images can be used for this use case. Aim for at least 1 image for exercise (2 would be ideal) with webp format and white background so they are lightweight.

[b3nj5m1n]

But yeah I agree: cloud models offer the best quality; just be sure their generated images can be used for this use case. Aim for at least 1 image for exercise (2 would be ideal) with webp format and white background so they are lightweight.

I've actually already added a few in 5aecc1b if you want to take a look, they seem to be around the same size as existing images.