Raspberry pi face detection and recognition with face rotational invariance based on Haar cascade and local binary pattern approach
If you have just set up the raspberry and the camera. Then please follow this tutorials for installing opencv and python. That a very good starting point.
Environment:
- Raspberry Pi 3 + Raspberry pi camera V2
- OpenCV 3.1.0
- Python 2.7.9
Implement:
- Open the dataset folder for sample images, the format in this dataset folder is User.[id].[index]. For example: User.1.43 is User with id=1 and the index of the image is 43. You should capture the user image in different expression, possition, lightning condition like in the dataset.
P/S: Open changeImageName folder, you can put all image of a user to image_here folder to change the image name format. Open changename.py and edit the image name format as you want then put the images back to dataset folder.
2. After colletecting the dataset please run: python trainer.py to train the images
3. After completing the training, please run: python FaceRecognizer.py to fire up the camera
#Face not in frontal position
#Face not in frontal position
References:
[1] Dasgupta, Anirban, et al. "A Vision-Based System for Monitoring the Loss of Attention in Automotive Drivers." IEEE Trans. Intelligent Transportation Systems 14.4 (2013): 1825-1838
[2] Ahonen, Timo, Abdenour Hadid, and Matti Pietikainen. "Face description with local binary patterns: Application to face recognition." IEEE transactions on pattern analysis and machine intelligence 28.12 (2006): 2037-2041.