A Deep Learning/Image Processing project created for Conant Hack Club.
Check out a Medium article explaining the process!
- Python 3.6
numpy
Keras 2.1.5
cv2 (OpenCV)
tensorflow 1.8
Download a zipped version of the project and locate the directory. Then, using an IDE or command line, run the test.py file. A window should now pop up with the program running on it.
- Take an orange highlighter, and run test.py
- Draw a letter in front of you
- Press E to erase, Press
- Press D when you are done
- The letter you drew will appear on the screen
- Rinse and repeat
- If you want to change the highlighter color, find the HSV of the chosen highlighter and update the mask values in test.py
- First, the computer tracks your orange highlighter based on its colors and draws its path. It does this by using contours and masks.
- Next, the computer crops out the path of the highlighter using contours, resizes and reformats the colors, and feeds it into a convolutional neural network
- The CNN spits out a letter, A-Z
- Prior to all of this, the CNN was trained using the EMNIST dataset