InternityFoundation / FaceRecognition-TransferLearning

FaceRecognition-TransferLearning

The whole project follows 3 steps of creating a Machine Learning Code i.e - a) Data Preparation . b) Model Training . c) Model Prediction .

Data Preparation (Take my photos.ipynb)

Your system must have a webcam , installed Python 3.x and opencv-python to run the code . If youwant to train the model to predict 5 people , then take 60 photos of each preson and save them in a particular folder . Create a Training Dataset and Validation Dataset and put these two folders in a single folder . Dataset is ready . Now upload this Dataset on Google Drive .

Model Training (Face_recog_Vgg16.ipynb)

Combining the models and printing the summary

Change the num_classes=5 as required (if your dataset has 6 classes then change to num_classes=6) .

Importing dataset and using Image generator for Augmentation

Change the train_data_dir and val_data_dir and specify the path of your dataset on the google drive.

Training the Model

Change the path where you want to save your model with model_name.h5 nb_train_samples = (total no. of images in your train dataset) nb_validation_samples = (total no. of images in your test dataset)

Model Prediction (Vgg16_prediction.ipynb)

classifier = load_model('path to model/model_name.h5') Update the face_dict and the face_dict_n with the names of your classes that you are predicting and the class names should be same as the folder names given in the prediction set . input_im = getRandomImage("path of prediction folder")

Model Training and Prediction(INTERNITY_face_detector_RAHUL.ipynb)

Training

Upload the zip file of the training_data and validation data(testing_data) and unzip them in the code . Change the num_classes=5 as required (if your dataset has 6 classes then change to num_classes=6)

Prediction

Change the path of prediction_data (data should have some images of each class in a separate folder put in a simgle folder) input_im = getRandomImage("path of prediction data folder") Update the face_dict and the face_dict_n with the names of your classes that you are predicting and the class names should be same as the folder names given in the prediction set .