Note that the above is still in its experimental changes as the algorithm lacks stability. Update(2/7/2021) The algorithm, after implementing Background Subtraction, in place of Backprojection, has become much more stable.
Important Things To Note: The hands exposure to light and color relative to the background is important when atttempting background subtraction. Thus, to ensure that algorithm is run as intended, run in an environment where hand is visibly illuminated and a plain background is present.
The code above is an attempt to experiment with skin detection with different colorspaces through Image BackProjection with the OpenCV library for Python. (Update) Upon researching further methods, the current implementation now takes advantage of a method called Background Subtraction, which will be further illustrated below.
This is transformed into two applications, finger counting and mouse tracking with OpenCV. There will be instructions provided below on how to use these algorithms for your own experimental purposes.
Ensure that you have the latest versions of the following packages:
- NumPy
- OpenCV
- Pickle
- Pyautogui
To install these, perform the following:
pip install numpy
pip install opencv-python
pip install pickle
pip install pyautoguiAs skin and lighting may differ depending on usage, the code utilises backprojection to tackle this variance. More information can be found here: https://docs.opencv.org/master/dc/df6/tutorial_py_histogram_backprojection.html
First run test_colors.py and attempt to fit hand within rectangle, as shown in the image below:
Upon pressing, two new files should be generated: model_hist_plot.png and model_hist.pkl. The model_hist_plot image shows the color distribution of the HSV colorspace (Hue, Saturation, Value).
Next we can either run, count_fingers.py or mouse_tracker.py. The image below demonstrates the use of count_fingers.py.
Lastly, please note that performance may differ heavily depending on background environment and light. Any new suggestions for other algorithmic implementations are always welcome. Improvements are being worked on.
Insert Information Here About Background Subtration Method
Have fun!