Fully Connected Artificial Neural Network based approach for Face Detection


Sachin Bansal , Galgotias College of Engineering and Technology; Shubham Varshney, Galgotias College of Engineering and Technology


Face Detection, Pattern Recognition, Computer Vision, Artificial Neural Networks, Machine Learning


In this work, A Neural Network based approach is presented to detect upright frontal faces in an image. The Neural Network used is a fully connected network with 3 layers. This network examines 20x20 sized windows in an image, and decides whether it contains a face or not. For the negative face examples, we use bootstrap method. After an initial small dataset of non-face examples, we add false detections to the training set as the training progresses. Simple heuristics, like merging overlapping and arbitrating between multiple neural networks can further increase the accuracy of the model.

Other Details

Paper ID: IJSRDV4I30219
Published in: Volume : 4, Issue : 3
Publication Date: 01/06/2016
Page(s): 364-366

Article Preview

Download Article