Face Detection Using VGG16 through Transfer Learning

Original article was published on Deep Learning on Medium

Face Detection Using VGG16 through Transfer Learning

Using the Deep Learning Approach

Task Description:

  • To create a Face Detection Model
  • By using the Transfer Learning concept
  • Can Use any of the Architectures like ResNet, VGG, Inception, MobileNet etc.,

My Project Components:

Here we will be using the environment especially created for Deep Learning using conda ( that has Keras, Numpy, Open CV etc., modules installed )

Data Set:
The data set we will be using in this project is the “ 14 Celebrity Faces Data set “ and can be downloaded from Kaggle

Data Set for this Face Detection Model


1.The code is written in Jupyter Notebook

2. Create a new Python Notebook

3. Code for the respective project

Facts of this Model:

  • This similar way of creating the model can also be used for the other Architecture
  • This model was tested on the data set ( from online ) but performed the task of Face Detection
  • This model was trained for some large time due to the architecture and no. of parameters we have used ( as VGG consumes high storage when compared to other architecture and computational problems)
  • Since the data set consists of around 200+ images for training and 70+ images for testing ( which is very less for achieving high accuracy and desired results )
  • The accuracy of this model ( with the above training is around ) 75–85% which is pretty high for the above conditions

Ways we can use this Project:

  • This project can be also used for custom face detection ( Using the transfer Learning that was the motive we want to achieve )
  • We can also attach the face authentication of our device with this model where our input will be our face