In this project, I developed a CNN that could draw bounding boxes around any face given in a image - a facial detection algorithm.

I programmed and trained the AI in Python using Tensorflow. The algorithm would parse the image into sections which would be evaluated by the CNN individually. For each “section“ of the image, the CNN would classify that spot as a “face“ or “not a face“. This way, the algorithm would be able to draw bounding boxes around all faces given in an image. After experimenting withh many architectures for the CNN, I settled on a 12 layer CNN with alternating layers of 2D convolutional layers and 2D maxpooling layers followed by a series of deep layers.

The article I wrote on CNNs ended up getting published in a medium publication named “Analytics Vidhya“

The final version of the AI algorithm ended up being 197 lines of code in total. In the totality of the project, I wrote ~ 500 lines of code.