Computer

Vision
Computer Vision
This course is designed to equip you with a deep understanding of computer vision, by mastering
image processing, neural networks, and advanced models, you will gain the ability to analyze,
interpret, and manipulate visual data.

Learning Objectives
Develop a strong foundation in image processing and deep learning for computer
vision.
Understand and implement Convolutional Neural Networks (CNN) for image
classification.
Apply advanced object detection techniques like YOLO and Faster RCNN.
Explore vision transformers and how attention mechanisms enhance computer vision
tasks.
Master segmentation methods for precise image analysis using models like Unet and
DeepLab.
Implement and evaluate real-time object tracking systems with algorithms like SORT
and DeepSORT.

Computer Vision Page 2

Course Information

Prerequisites

The Computer Vision course is an advanced program and requires previous competence in
the following areas:

Programming Fundamentals : Proficiency in Python, especially with libraries like NumPy,

Pandas, and OpenCV.

Basic Machine Learning Concepts : Familiarity with supervised and unsupervised

learning, and experience in training foundational machine learning models.

Neural Network Basics : A foundational understanding of neural networks, covering topics

like neurons, layers, activation functions, and backpropagation, will help you follow advanced
concepts taught in the course.

The course is designed to be completed over a duration of approximately four months, allowing for
a thorough exploration of advanced computer vision concepts while providing ample time for
hands-on practice and application.

Estimated Time Required Skill Level

4 months 6hrs/week* Begineer+

Computer Vision Page 3

Course Instructors
Our course is led by two seasoned experts in AI, machine learning, and computer vision, bringing
together a wealth of real-world experience and deep technical knowledge to guide your learning
journey.

Krish Naik : Krish Naik is a seasoned AI engineer with over 15 years of experience in machine
learning, deep learning, and computer vision. His expertise includes advanced generative AI
techniques, model development, and implementation of AI solutions across diverse use cases.
Krish’s extensive industry background ensures learners gain a grounded understanding of cutting-
edge ML and AI technologies.
Monal Kumar : An expert in computer vision and a full-stack data scientist, Monal Kumar brings
extensive industry experience, particularly in live deployments of AI solutions. With a strong
background in end-to-end project development and specialization in computer vision, Monal’s
insights will help you navigate both the technical and practical aspects of AI-driven visual
solutions.

Monal Kumar Krish Naik

Data Scientist Chief AI Engineer
Linkedin Linkedin

Computer Vision Page 4

 Module 1

Computer Vision Introduction

This module introduces the foundational concepts of computer vision and image processing. You'll
explore the basics of deep learning and how it applies to image data, including the structure of
images, color models, and key image transformations like scaling, cropping, and rotating. By the
end of this module, you'll understand how to manipulate images and apply initial classification
techniques, setting the stage for more advanced methods in later modules.

Topics

Foundations of Image Processing Understanding Pixels, Image Types, EXIF

Color Models Color Models, Color Thresholding, Image

Classification

Image Manipulation and Image Scaling, Aspect Ratio, Crop, Image

Transformation Flip, Rotate

Image Features Contours, Contours Processing

Computer Vision Page 5

Module 2

DL - Computer Vision I
This module dives into the fundamentals of deep learning applied to computer vision, focusing on
neural networks. You’ll learn about the essential components of neural networks, including
neurons, layers, activation functions, and backpropagation. The module also covers the basic
techniques for image classification, helping you understand how deep learning can be used to
recognize simple patterns in visual data through digit recognition tasks with a vanilla neural
network.

Topics

Deep Learning Concepts Types of Learning, Understanding Image

Data Variation: occlusion, scale variation,
illumination, noise, background & other

Neural Network Fundamentals Components of Neural Network

Core Mechanisms of Neural Networks Activation Function, Loss Function,

Optimizer, Forward Propagation,
Backpropagation, Learning Rate

Hands-on Practice Digit Recognition with Vanilla Neural

Network

Computer Vision Page 6

Module 3

DL - Computer Vision II
In this module, you’ll delve into Convolutional Neural Networks (CNN), the cornerstone of modern
computer vision. You'll explore why CNNs outperform traditional neural networks in image tasks
and learn about key components like filters, pooling, and dense layers. By the end, you'll have a
solid understanding of CNN architecture and its applications, and you'll be able to implement your
own CNN model for basic image recognition tasks using architectures like LeNet.

Topics

Introduction to CNNs Convolution Neural Network, Why CNN

is Better than ANN, Components of CNN,
Input Data

Core CNN Operations Convolution Layer, Convolve Function,

Filters (Kernels), Kernel Size, Stride,
Padding, Feature Map, Channels

Activation and Pooling Activation Function, Why to Use

Activation Function, Pooling Layer, Max
+ AVG Pooling, 1x1 Convolution

Network Architecture Flattening, Fully Connected Layer

(Dense Layer), Dropout, Batch
Normalization, Softmax

Computer Vision Page 7

Module 3

DL - Computer Vision II

Topics

Training Mechanisms Loss Function, Optimizer, Forward

Propagation, Backpropagation

Output and Predictions Output

Computer Vision Page 8

Module 4

DL - Computer Vision III

This module focuses on advanced CNN architectures that have shaped the field of deep learning
in computer vision. You’ll explore key models like AlexNet, VGGNet, GoogLeNet, ResNet, and
MobileNet, which revolutionized the way visual tasks are tackled. Understanding these
architectures will give you the knowledge to choose and implement state-of-the-art models for
various image classification challenges, optimizing for accuracy, efficiency, and scale.

Topics

Early CNN Architectures LeNet, AlexNet, VGGNet

Advanced and Deeper Networks ResNet, Inception-v3

Efficient and Mobile-Friendly DenseNet, MobileNet

Architectures

Computer Vision Page 9

Module 5

DL - Computer Vision IV - Computer

Vision with Attention
This module introduces the concept of attention mechanisms in computer vision, particularly
through Vision Transformers (ViT). You'll learn how transformers, originally designed for NLP, are
transforming vision tasks by capturing relationships across an entire image. The module also
covers ConvNeXt, a hybrid approach combining CNNs and transformers, providing you with
cutting-edge techniques to enhance model performance in visual tasks.

Topics

Introduction to Transformers in Vision Why Use Transformers to Solve Vision

Tasks, Vision Transformers (ViT)

Input Processing and Representation Input Representation, Positional

Encoding, Class Token

Core Transformer Mechanisms Multi-Head Self-Attention, Feed-

Forward Network, Layer Normalization,
Residual Connections

Model Architecture Encoders, Output Head

Computer Vision Page 10

Module 6

DL - Computer Vision V - Object

Detection
In this module, you will learn the intricacies of object detection, a key technique for identifying
objects within an image. You’ll explore both two-stage (e.g., Faster RCNN) and single-stage (e.g.,
YOLO) detectors, along with techniques like Region Proposal Networks (RPN) and anchor boxes.
By the end, you’ll have the skills to build and fine-tune object detection models for real-world
applications, such as autonomous vehicles and security systems.

Topics

Introduction to Object Detection What is Object Detection, Classification,

Regression

Core Components of Object Selective Search, Region Proposal

Detection Network, CNN - Feature Extractor, Pre-
trained Backbones, Feature Pyramid
Network, RoI Pooling, Anchor Boxes,
Bounding Box Regression &
Classification Head

Object Detection Algorithms Two-Stage Detectors : RCNN, Fast

RCNN, Faster RCNN.
Single-Stage Detectors : YOLO, Object
Detection using YOLOv5 & YOLOv11

Advanced Detection Methods Non-Maximum Suppression, Advanced

Loss Functions

Computer Vision Page 11

Module 6

DL - Computer Vision V - Object

Detection

Topics

Hands-On Creating Our Own Object Detection

Algorithm

Computer Vision Page 12

Module 7

DL - Computer Vision VI - Segmentation

This module explores segmentation, which focuses on classifying individual pixels in an image to
distinguish objects and regions. You’ll dive into semantic segmentation using models like Unet and
DeepLab, and explore instance segmentation techniques such as MaskRCNN. Through hands-on
practice, you’ll learn to implement these techniques for tasks like medical imaging and scene
understanding, making your models capable of high-precision visual analysis.

Topics

Introduction to Segmentation What is Segmentation, Semantic

Segmentation, Instance Segmentation

Core Concepts in Segmentation Downsampling, Upsampling/Transposed

Convolution, Skip Connections, Atrous
Convolutions, Conditional Random
Fields, Loss Functions (Softmax with
Cross-Entropy, Dice Loss), Evaluation
Metrics

Popular Architectures & Framework Unet, DeepLab v3, MaskRCNN,

MMDetection

Practical Implementation Implementing Unet from Scratch,

Popular Datasets to Get Started

Computer Vision Page 13

Module 8

DL - Computer Vision VII - Object

Tracking
This module covers object tracking, a dynamic task in computer vision that focuses on following
objects as they move through video frames. You’ll explore methods like filter-based tracking, CNN-
based tracking, and advanced algorithms such as SORT and DeepSORT. By mastering object
tracking, you’ll be able to apply your skills in areas like surveillance, sports analytics, and
autonomous driving, where tracking objects in real time is crucial.

Topics

Introduction to Object Tracking What is Object Tracking

Tracking Methods Filter-Based Tracking, CNN-Based

Tracking

Key Algorithms in Object Tracking Kalman Filter, SORT, DeepSORT , Re-ID

Hands-On Using YOLO and ByteSort to Track

Objects

Computer Vision Page 14

Module 9

PRO Module - Generative Models for

Vision Applications
This module dives into generative AI models specialized for vision applications, including cutting-
edge tools such as CLIP, SAM2, Stable Diffusion, and CycleGAN. You’ll explore how these models
are used for tasks like text-to-image generation, segmentation, style transfer, and more. By
mastering these techniques, you’ll be able to create and manipulate visual content with precision
and apply these skills in fields like digital art, automated content creation, and synthetic data
generation for training other models.

Topics

Introduction to Generative AI Overview of Generative AI in Vision

Applications in Image Synthesis, Style
Transfer, and Segmentation

Key Models and Techniques CLIP, SAM, Stable Diffusion, CycleGAN

Hands-On Text-to-Image Generation for

Creative Media
Image Segmentation for Data
Labeling
Style Transfer and Domain
Adaptation for Synthetic Data

Computer Vision Page 15