Using python OpenCV module to train and recognize up to 5 faces and generate automated voice feedback
Creator.py --> Uses your webcam to take multiple pictures of your face and crops it and turns to grayscale to be analyzed later
Trainer.py --> Uses the saved pictures to start analyzing trends and picks up on differentiating features of each face to generate a .YML file
Detector.py --> Will Use the .YML file to run the actual face detection program. It will provide AUDIO FEEDBACK based on who it recognizes. It will also print the percent confidence of each of the detection. (You can change this in the detector.py prgram { variable called conf })
-
All 3 python files
-
2 folders with "dataSet" and trainer" name
-
A webcam set to default and good lighting conditions
STEP 1: Run creator.py and wait until picture taking process is completed
STEP 2: Run the trainer.py to analyze the JPG pictures and create YML file. IT will ask for a number, which is the ID of each of the 5 possible faces. Choose 1-5 accordingly and hit enter.
STEP 3 (OPTIONAL): save .MP3 files to the same folder with audio that you want to play for each unique face detected successfully
STEP 4: Edit detector.py to change the mouse click location to the trigger button in the alexa app to automate any smart home appliance!
STEP 5: Run the detector.py to begin facial analysis/detection