Human-Detection

Neural Network based Human Detection in 2D color images with HOG features.

Files uploaded :

1. CV_hog_file.py. - [HOG Feature computation of an input image]
2. Neural_network.py - [Customised Neural Network Classifier Ground-up Implementation]

A complete implementation of Histogram of Oriented Gradients (HOG) feature and a Three-layer Perceptron(Neural Network) for detecting human in images involving the following steps -

1. Normalisation of color to grayscale image.
2. Prewitt's Horizontal and Vertical Gradient Operation for Magnitude and Gradient Angle computation.
3. Normalised Computation of HOG feature.
4. Designing a 2-layered Perceptron for Supervised classification.
5. ReLU Activation function for neurons in the hidden layer and Sigmoid function for the output neuron with Backpropagation for training multi-layer Perceptrons.

Technologies used - Python : NumPy and OpenCV.