No training scripts and event models as well as inference scripts !

[amitpant7]

Hello,

I was going through the repo, I checked the examples. The demo and work surprised me. HPE on event data of this level is really astonishing with high accuracy. I wanted to replicate the work. Are those movenet examples of the examples for the move event net. Also, there isn't any training script as well as model and any discription about the data formats. Also it would be great if there was an example to inference your model on a single data instance.

Can you please update this?

Regards
Amit

[arrenglover]

Hi @amitpant7 ! Thanks for your interest. I'll ask @gaurvigoyal to give you some more information!

[gaurvigoyal]

Hey Amit! Thanks for your interest. They are in the example folder movenet (here). There are instructions and a docker for the environment as well. We also provide all of our own models discussed in the ablation study, to support replication.

But I saw that it's a bit stale, since we published it over a year ago. So, I'm checking things, and adding a python-only inference. I'll link the pull request here.

[amitpant7]

Thanks for the update. I will check the examples.

[gaurvigoyal]

Hey @amitpant7 , we've merged the pull request (link) with one more option for inference and better instructions.

We'd like to ensure others can also reach the right folder for this in the future, so I've tried to clarify this in the first Readme.md on this page. Would you mind having a look and telling me if it's clear?

[amitpant7]

Thanks, the inference is now easy to follow but I couldn't find any scripts to train or refine the networks .

[gaurvigoyal]

Hey @amitpant7,

The code here is only for inference. There is a separate repository with the training code. It's not as clean and as of now the instructions are outdated. If you think access to the training code would benefit your work, we can update that repository and link it to this one.

Let us know!

[amitpant7]

Hello @gaurvigoyal

Thank you for the update! I’d be happy to go through the training script. Additionally, if there are any specific documents outlining the training methods and processes used, I’d greatly appreciate the opportunity to go through those as well.

Looking forward to your response!