Tú puedes: an open source smart gym

Tú puedes will count and correct you exercises using computer vision. Tú puedes calculate you body pose and training equipment position using live computer vision, to count you repetitions, alert you on mistakes, track and motivate you to go further.

Install

conda create -n tupuedes310 python=3.10
conda activate tupuedes310
git clone git@github.com:verasativa/tupuedes.git
cd tupuedes 
pip install -e .

Usage

Usage like

tupuedes train

or

tupuedes train --source=3

where you source is 0 for default web cam (other ints for other webcams), or a network camera like http://192.168.1.155:8080/video

Sharing a camera from phone

• Android
  • IP Webcam dose the work nice

Philosophy

Tú puedes could be eventually run on mobile devices, eventually. But before product-market fit, we rather to iterate fast; that why:

• We manage video trough openCV from python
• we would like to our python be really pythonic (run ```import this`` at ipython/notebook cell)
• We provide a full pandas dataframe with structured data to exercise classes

Tú puedes would like to be open source and commercial, like:

• anaconda
• posthog

Tú puedes would like to be a feminist artifact:

• how this is empowering women? is always a good question
• proprietary software is kind of patriarchal
• best competition is against you self

dev goals

To públic access

• 1 naive working exercise
• home dir writting (mac/windows/linux)
• nice readme
  • demo gif
  • conda set up
  • windows / mac / linux
  • vision / press release
    • import this
    • feminist
• discord channel?
• setting metrics
• pass the source trough **kargs to super().init() in the fancy way at all pipeline child classes

good practice stuff (coming soon)

• train a pose detector (time series labeller) with sktime pipeline including exponential smoothing and detrend
• set a standard interfaces for body positions and/or body pode embedding
• set exercises model training as dvc experiments ref
• refactor window to pygame
  • set rectangles and minimal viable displays for:
    • repetitions
    • mode
    • badges
    • telemetry (bpm and other sensors)
    • debug/plot
  • home menu

Maybe late

• config manager
• run it on PinePhone2
• explore compiling to webassembly