peloton-bike-metrics-server

The code in this repository acquires Peloton® Bike metrics (such as cadence and output) by using Tesseract OCR to perform digit recognition on the projected screen from the Android table installed on the bike.
Metrics are exposed at /metrics REST endpoint so they can be consumed by a wide range of applications.

As an example, the code in this project can be used to integrate with zwack-bike to allow Peloton® Bike owners to take rides on Zwift.
Diagram of the project:

Requirements

Windows (tested on Windows 10 x64, Version 1803)
Python 3.x
Peloton® Bike that runs on Android

Installation

Install Python 3.x
Install Tesseract OCR should be installed on the machine (tested on Tesseract 4 for Win x64)

Clone repository, create virtual environment and install python dependencies:

git clone https://github.com/iaroslavn/peloton-bike-metrics-server.git
cd peloton-bike-metrics-server

virtualenv metrics-server-env
metrics-server-env\Scripts\activate.bat

pip install -r requirements.txt

Running the Peloton Bike Metrics Server

Allow Peloton® Android device to project its screen to your Windows monitor:
On Peloton® Bike navigate to Settings > Cast Screen and connect to your PC:
Start a ride session on Peloton® Bike:

Start the Peloton Bike Metrics Server:

cd peloton-bike-metrics-server

metrics-server-env\Scripts\activate.bat

set PATH=C:\Program Files\Tesseract-OCR\;%PATH%

python metrics-server.py

Bring the Connect window to the front
Open http://127.0.0.1:5000/metrics in your browser and verify that metrics are being returned. E.g.:
```
{"cadence": 85, "power": 102}
```
Check out the zwack-bike project if you want to consume these metrics by BLE device simulator and report them to an application such as Zwift.

Debug Flags

Use -d or --debug flag to start the server in the debug mode:

python metrics-server.py -d

Use -p or --port flag to change the default web server port:

python metrics-server.py --port 5001

Known limitations

The Connect window with the projected Peloton® Bike screen have to be visible so the metric numbers can be recognized.
Depending on your screen you might have to adjust rectangle coordinates in get_power_and_cadence_imgs function. Turn on the debug mode (-d flag) to see what's been recognized by the OCR process.

Donations

If you like this project, please consider doing the following:

Starring on GitHub
Buying me a coffee

iaroslavn / peloton-bike-metrics-server

readme