The following python scripts allow one to scalably mine biometric data (e.g. Heart Rate + Power) from Zwift biking videos. Emotional facial expression analysis was extracted utilizing iMotions (Version 8.0).
Code Dependencies: Tesseract 4.1.1, ffmpeg 4.2.2
ffmpeg.py - Provides basic functionality including the ability to slice videos, change the frame rate, concatenate two videos together, crop the video, and, lastly, split the video into its constituent frames.
tesseract_flow.py - Handles image pre-processing and runs tesseract OCR ("psm 7 outputbase digits") on a directory full of frames. Before running OCR, images were cropped and subsequently subject to sharpening and binarization. See tess_data for raw tesseract results.
data_analysis.py - Cleans data and removes outliers (incorrect results) from tesseract predictions. Visualizes resulting data.