Dependencies

• Firefox
• BeautifulSoup
• selenium (with geckodriver.exe in PATH)
• opencv-python (using 4.5.1) with GUI
• youtube-dl
• ffmpeg
• pytesseract — must be with 14MB eng.traineddata (replace the 4MB one)

TODO List

• Account for missing checklist more dynamically
  • Add log to keep track of previous detections and calculate time since
• Use line detection instead of contours to detect the box is on screen
• Verify initial crop and min/max area values based on real data
• Add more comprehensive crop for entire checklist (can use cv2.vconcat)
• Filter to check for duplicate crops (use cv2.compareHist with thresh2?)
  • Avoids running through entire thing every time
• Split checklist using cv2.reduce to read every line separately
  • Can get sequence number
• Figure out some way to share the info openly
• Add a flag to delete captured images after
• Add timeouts, exceptions & cleaner prints
• Add coordinate sorting to crop_img_from_contours

Known Issues

• Sometimes get_stream_screenshot takes excessively long; this is an ffmpeg issue that I haven't figured out yet
• There may be some issues running from a batch file due to working directories and where the images are being saved. Have to look into this more.

Changelog

• 2021-02-23: Added fix for case where findContours returns redundant vertices and made a couple things more streamlined