Mirror of forum.nim-lang.org

5297 :: multi-platform time tracker with ML use

• dponyatov (orginal) [2019-10-07T08:17:56+02:00] view original

  What packages and sample code projects can I use to write a multi-platform time tracker with ML use?

  What I need:

  • take a screenshot every Nth second (one in a minute is affordable I think)
  • encode it into daily or weekly video ring buffer (for logging)
  • OCR screenshot for text analysis: what software in active runs, what project, etc
  • task process list snapshot (what software runs in background in case of compiling etc)
  • feed OCRed text, process list etc into ML engine to detect what projects and tasks under active development, and what is non-target user activities

  I mostly target it for myself to be more effective (background automated work tracking) and not going to distribute it for employer spying. But it will be MIT license, so nobody is not limited to do anything with it, so don't help me if you hate the big brother kickass software.

---

• mratsim (orginal) [2019-10-08T15:10:51+02:00] view original

  For OCR you will need OpenCV compiled with Tesseract support.

  See tutorial here, especially using the C++ API: https://www.learnopencv.com/deep-learning-based-text-recognition-ocr-using-tesseract-and-opencv/

  I think the deep learning API of OpenCV is only exposed through C++ while the opencv package from Nim (at https://github.com/dom96/nim-opencv) only exposes the C API so you might need to regenerate your own wrapper (probably nimterop could help)

---