Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 6

Artificial Intelligence with Python – Deep Neural Networks

Section 1: Object Detection and Tracking

1.1 The Course Overview

This video will give you an overview about the course.

1.2 Installing OpenCV

In this video we will be installing OpenCV for multiple operating systems.

o Go through the links for installing OpenCV for different operating systems

1.3 Frame Differencing

Frame differencing is one of the simplest techniques that can be used to identify the moving parts in a
video. We will learn more about it in this video.

o Compute the frame differences

o Scale and resize the frames

o Iterate and update the frame variables

1.4 Tracking Objects Using Colorspaces

In this video we will make use of colorspace to build a robust tracker also track it using the same.

o Grab current frame and define scaling factor to resize images

o Convert image to HSV colorspaces

o Compute bitwise-AND between mask and original image, display input and output

1.5 Object Tracking Using Background Subtraction

Background subtraction performs well for moving objects in a static background. Let’s see how it works
in this video.

o Grab the current frame, resize it and return it

o Define the background subtractor object

o Compute the mask and convert it into RGB, display input and output
1.6 Building an Object Tracker Using the CAMShift Algorithm

Color space based tracking allows us to track colored objects, but we have to define the color first. This
seems restrictive! Let us see how we can select an object in a live video and then have a tracker that can
track it.

o Capture the current frame and track the rectangular selection and object

o Convert the color space from RGB to HSV and extract the regions of interest

o Compute the histogram backprojection and bitwise-AND of histogram and mask

o Apply CAMShift algorithm

1.7 Optical Flow Based Tracking

Optical flow is a very popular technique used in computer vision. It uses image feature points to track an
object. Let’s learn more about it in this video

o Define tracking parameters and convert frame from RGB to grayscale

o Compute the optical flow based on previous and current images

o Iterate through good features and circle them. Update the features

o Define main function and start tracking

1.8 Face Detection and Tracking

Face detection refers to detecting the location of a face in a given image. In this video, we will see how
to detect face and eyes.

o Use Haar cascades to detect object

o Use integral for images to extract features

o Perform eye detection and tracking


Section 2: Artificial Neural Networks

2.1 Introduction to Artificial Neural Networks

An artificial neural network is a model designed to simulate the learning process of the human brain.
This video is a brief introduction to them.

o Build the neural network

o Train the neural network

2.2 Building a Perceptron Based Classifier

A Perceptron is the building block of an artificial neural network. It is a single neuron that takes inputs,
performs computation on them, and then produces an output. In this video, we will build a classifier
using perceptron

o Separate the text into datapoints and labels

o Define maximum and minimum datapoints

o Train the perceptron and plot the training progress

2.3 Constructing Single and Multilayer Neural Networks

A perceptron is not enough to perform a task. So we will construct a single layer neural network and also
improvise it by creating a multilayer network.

o Create a single layer neural network, train it, and run it

o Construct a multilayer neural network with two hidden layers, train it and run it

o Compare both

2.4 Building a Vector Quantizer

Vector Quantization is a quantization technique where the input data is represented by a fixed number
of representative points. We will build a vector quantizer in this video.

o Define a neural network and train it. Create a grid of points

o Extract the classes and grids corresponding to these classes

o Plot the outputs

2.5 Analyzing Sequential Data Using Recurrent Neural Networks

We have been dealing with static data so far. Artificial neural networks are good at building models for
sequential data too. For that, we have recurrent neural networks.
o Generate waveform and create varying amplitude for it

o Create a recurrent neural network. Train it and run the data

o Test the performance

2.6 Visualizing Characters in an Optical Character Recognition Database

Optical Character Recognition is the process of recognizing handwritten characters in images. Before we
jump into the model, we need to be familiar with the dataset. Let’s do that in this video.

o Define visualization and other parameters

o Reshape the 1D array into a 2D image

o Display the image

2.7 Building an Optical Character Recognition Engine

Now that we have learned how to work with the OCR dataset, let's build an optical character recognition
system using artificial neural networks in this video.

o Now that we have learned how to work with the OCR dataset, let's build an optical
character recognition system using artificial neural networks in this video.

o Extract the current label and character and append it to the list

o Convert lists into numpy arrays. Create a feedforward neural network and train it.

o Predict the output for test data

Section 3: Reinforcement Learning

3.1 What Is Reinforcement Learning?

This video will be an introduction to reinforcement learning.

o Understand reinforcement learning

o Look at real world examples and building blocks of reinforcement learning

3.2 Creating an Environment

Environment is an important component of reinforcement learning. We first need to create an


environment for the testing and training.

o Specify the type of environment you want to run

o Parse the input arguments and create a mapping


3.3 Building a Learning Agent

The learning agent will learn how to achieve a goal. So it is important to know how to build one.

o Parse the input arguments and map them

o Start iterating by resetting the environment

o Extract consequences of taking current action

Section 4: Deep Learning with Convolutional Neural Networks

4.1 What are Convolutional Neural Networks?

In this video, we will just see what convolutional neural networks are and understand the basic working.

o Study the architecture of CNNs

o Look at the types of layers in CNN

4.2 Building a Perceptron-Based Linear Regressor

In this video, we will build a linear regression model using neural network approach.

o Generate some datapoints and define the model

o Generate some noise. Separate data into input and output variables

o Define the gradient descent optimizer and specify the loss function

o Start the TensorFlow session and training process

4.3 Building an Image Classifier Using a Single Layer Neural Network

In this video, we will build an image classifier to classify handwritten digits using single neural network..

o Extract the MNIST image data

o Convert data into a single dimensional array

o Create a single layer neural network and train it with the variables

4.4 Building an Image Classifier Using a Convolutional Neural Network

The image classifier in the previous video didn't perform well. Hence we will build a better one using
CNN.

o Create weights and biases. Perform 2D convolution

o Extract the MNIST image data. Reshape x into 4D tensor


o Apply ReLU and start training and run optimizer

You might also like