face-recognisation-by-transfer-learning

In this project, I have created a Face-Recognition model using the concept of Feature Tuning.

Step 1: We start of by collecting our dataset. For this, I have used Haarcascade FrontalFace. I have collected 200 images of mine & my friend for training the model & 100 images each for testing the model. You can use the following code to collect the images and prepare the dataset.

I have used the same block of code multiple times to collect all the training & testing images of me & my friend. You can collect images of more people as per your requirement. For more details check the code

Step 2: Now, we import pre-created MobileNet model from keras.applications. We freeze the already trained layers by layer.trainable= False.

Step 3:  We add layers as per our requirement. Here, I have used Softmax activation function.

Step 4: Next, we load our dataset. We have used the augmentation technique to increase our dataset since the size of original dataset is too small for a good accuracy.

Step 5: Now, we begin training our model.

The model has been effectively trained and ready to use. You can use this model for prediction.

In this model, I got 99% accuracy, because the data was very less.

Step 6: Now, I have loaded the created model for prediction, and predicted mine and my friend face.

The output of predicted model:

github link:https://github.com/Anuddeeph/FaceRecognisation-Mobilenet-.git

thanks to Vimal Daga sir, for guiding us.

#artificialintelligence #facedetection #mlops #project #python  #devops #github #vimaldaga #git