trained pose-estimation weights

Hi rUv,

I'm building an elderly-care monitoring product on top of your WiFi-DensePose (RuView) project — fall detection, breathing, inactivity alerts, using 3× ESP32 + 2.4GHz, camera-free.

I've already: forked & synced to latest (a76adc3c), passed all 4375 tests locally, ran the simulated sensing-server, and studied the training pipeline (MM-Fi + tch-backend).

Request:

1. Do you have any usable trained pose-estimation weights (.onnx/.rvf/.safetensors)?
2. If not, could you share training hyperparameters / checkpoint config?
3. Any MM-Fi acquisition/preprocessing scripts you can point me to?
4. Does the Cognitum commercial version offer trained models/data, and at what price?

I'll respect the MIT/Apache-2.0 license, and I'm happy to give back (feedback, test data, code contributions). Happy to share my use case if useful to your product direction.

Thanks for the great open-source work!

Start with the training pipeline and tch-backend references mentioned in the issue, along with the simulated sensing-server context. Done means documenting whether trained weights, hyperparameters, checkpoint configuration, and MM-Fi acquisition or preprocessing scripts are available, plus an answer about the commercial offering.