Based on my recent lessons with Jeff Heaton on Deep Learning, I attempted to extend the power of Yolov3 from detecting pre-defined objects, from still images / videos to computer desktop display. This tool is more versatile as users can pull up different items on screen, browser various webpages, and see the object detection acting in live environment. My work is inspired and revised from this lesson.

I have to note that at my workstation setup, I have my laptop and x2 slave screens. Therefore it is easy enough for me to open up various items on one screen, while another OpenCV pop-up window displays on a second screen where Yolov3 object detection takes place.

It is only the first attempt of extending Yolov3's application, and the main intention comes from my workplace inspiration that when I need to explain an idea to my audience, it will help elaborate much easier with the use of Yolo's object detection functionality, as Yolo essentially supplement my presentation with on-screen annotations.

Major drawback so far is the serious display lag, guess it will be better with higher CPU PC. Regardless, it is still fun enough to put this tool into practice. Next step I envision is to train customized objects (e.g. by transfer learning) so that Yolo is able to pick up additional customized items other than the pre-defined ones.

Feel free to comment and let me know what you think. I have also linked this repository to my Medium blog, as I believe sharing of experience is vital to keep the ML community vibrant.

Feel free to fork and enjoy! :)