Craig Mouser made a website tool to extract telemetry from video files. It looks like you can upload a video file, select the regions on screen which contain the telemetry you want to extract, and it will OCR each frame and output a CSV file.
Announcement and explanation on Twitter:
https://x.com/mouser58907/status/1819483839535714648
I made a tool to extract telemetry from video files. I’ve seen too many people do it manually and thought it seemed like a fun challenge to build. I gave myself one day and It works pretty well. Let me know what you think! https://telemetry.mouserlabs.com
https://x.com/mouser58907/status/1819484432119611732
Worth noting it is all done locally and for now requires you to have the video locally.
https://x.com/RyanHansenSpace/status/1819495679875772812
Have you tried altering the video file to drop the background much darker and boosting the text to see if that helps the text recognition?
https://x.com/mouser58907/status/1819498167928533473
Yeah, that’s what I was attempting to do with the threshold slider. Tesseract just isn’t that great. If I had more time to work on it I would have done something with a backend and deep learning. This blog made me give up pretty quick on trying to improve it. https://blog.roboflow.com/best-ocr-models-text-recognition/
Seems like a pretty cool and useful tool!
This is awesome! Can’t wait to try it out. Over the years I’ve seen various tools, some of them paid, and of very different quality levels.
Usually I would revert to just manual data entry by making life easier for myself and just looking at every 5 seconds or so. Usually works well enough, but I am always left wanting to see the finer details.
Looking forward to testing this out :)