Course Syllabus

Module 1: Introduction to Computer Vision & Tools (Week 1)

This module builds the foundation of computer vision and sets up the tools. We introduce what CV is and survey key applications. Students learn the basics of digital images (pixels, color channels, formats) and the CV workflow. The session covers setting up the development environment: installing Python, C++, and OpenCV (with bindings for both languages). Learners write simple programs to load, display, and save images/videos using OpenCV in Python or C++. By the end of this module, participants will be able to describe the goals of CV, differentiate between prototyping (Python) and performance (C++) development, and write basic code to read and visualize image data.

Module 2: Fundamentals of Image Processing (Week 2)

This module covers core image processing techniques using OpenCV. Topics include color space conversions (RGB, grayscale, HSV), pixel-wise operations (brightness/contrast), histograms and thresholding, and filtering techniques. Students practice converting images between color models and computing image histograms. They apply filters such as Gaussian blur, sharpening kernels, and edge detectors (e.g. Sobel/Canny). Binary processing (thresholding, erosion, dilation) is introduced for basic segmentation. In hands-on labs, learners implement operations like histogram equalization and simple color segmentation. By the end, students will be able to apply image filters to enhance or segment images and explain how histograms and color thresholds are used in vision tasks.

Module 3: Feature Detection and Geometric Transforms (Week 3)

This module introduces feature detection and geometric transformations. Topics include detecting edges (Canny) and corners (Harris/Shi-Tomasi), as well as the Hough Transform for detecting lines and circles. We cover image transformations such as scaling, rotation, affine and perspective (homography) warping. Feature detection algorithms (SIFT, SURF, ORB) and feature matching using descriptors are also taught. Students learn to use RANSAC to match points between images and compute homography for tasks like image alignment or panorama stitching. In labs, students apply these methods to align images and detect shapes. By the end of this module, participants will be able to extract and match keypoints, perform affine/perspective transformations, and use feature matches to infer camera motion or create panoramas.

Module 4: Video Processing and Tracking (Week 4)

This module extends CV to video streams. Topics include video capture (from files or camera), frame-by-frame processing, and background subtraction for motion detection. We introduce motion estimation via optical flow (Lucas–Kanade) and teach object tracking methods (MeanShift/CAMShift, KCF, CSRT). Students implement a basic motion detector and track objects across frames. Lab exercises involve stabilizing shaky video and tracking a moving object in real time. By the end, learners will be capable of capturing and processing live video, detecting moving regions, and using tracking algorithms to follow objects in a video sequence.

Module 5: Classical Machine Learning for Vision (Week 5)

This module covers traditional machine learning techniques applied to images. Topics include feature-based classification and template/detector methods. Students learn to extract features (e.g. Histogram of Oriented Gradients, HoG) and train classifiers (SVM, KNN) for tasks like digit or face recognition. The module also introduces Haar-cascade classifiers for object detection (e.g. face detection) using OpenCV’s built-in cascades. In practical labs, participants train a simple image classifier and run a pre-trained Haar detector on video. By the end, students will understand how to use feature descriptors and classifiers for vision tasks, and will be able to apply a trained SVM or Haar cascade to classify or detect objects in images.

Module 6: Deep Learning Basics for Vision (Week 6)

This module dives into deep learning fundamentals for vision. We introduce neural networks and specifically Convolutional Neural Networks (CNNs) – explaining convolutional layers, pooling, and fully-connected layers. Students get hands-on with a framework (PyTorch or TensorFlow) to build and train a CNN for image classification. For example, they might train on MNIST or CIFAR-10 datasets. Topics include activation functions, loss functions, backpropagation, and training techniques (batching, optimization). As noted in industry, frameworks like PyTorch have made building advanced vision models much easier. Labs include implementing a simple CNN from scratch and evaluating its accuracy. By the end of this module, participants will be able to design a basic CNN architecture, train it on image data, and analyze its performance on classification tasks.

Module 7: Object Detection and Segmentation (Deep Learning) (Week 7)

This module covers advanced deep learning techniques for detection and segmentation. First, we introduce object detection models such as YOLO (You Only Look Once) and SSD (Single Shot Multibox Detector), and how to use OpenCV’s DNN module or PyTorch models to run them. For example, learners will use a pre-trained YOLOv5/YOLOv7 model to detect multiple objects in images. We also cover region-based detectors (Faster R-CNN) and instance segmentation (Mask R-CNN). Topics include transfer learning (fine-tuning on a custom dataset) and evaluation metrics (precision/recall, mAP). Hands-on labs involve running inference with a YOLO or Mask R-CNN model on sample data, and optionally training a small custom detector. In summary, students will learn to apply state-of-the-art models for real-time object detection and semantic/instance segmentation.

Module 8: Industrial Applications & Deployment (Week 8)

The final module ties everything together with real-world applications and deployment. We survey industry case studies: for example, how machine vision is used in manufacturing for defect inspection and automation, or in agriculture for sorting produce. Students learn about performance and optimization (using GPUs, batch processing) and explore hardware deployment. A special topic is deploying models on embedded platforms like the NVIDIA Jetson (Nano or Xavier). For instance, configuring a Jetson Nano with OpenCV and TensorFlow to run a vision model is a common embedded CV taskIn labs, participants practice exporting a trained model and running inference on a Jetson or similar device (or simulate this in class). By the end of this module, students will understand how to package and deploy computer vision systems, and will appreciate how CV pipelines are used in areas like robotics, smart manufacturing, and autonomous vehicles.

Key Features

• Hands-On Labs: Each week includes interactive coding labs using real images and videos. Participants write Python and/or C++ code with OpenCV, PyTorch, and TensorFlow to apply concepts immediately.
• Projects: The course culminates in guided projects (e.g. an object detection system or image classifier) that integrate learned techniques. A capstone project simulating an industrial use case (e.g. quality inspection or autonomous navigation) allows students to apply end-to-end pipelines.
• Assessments: Quizzes and short assignments after each module reinforce theoretical concepts (e.g. filter design, algorithm steps). Practical exercises (e.g. implementing a tracking algorithm) evaluate skills.
• Collaboration: Small group discussions and code reviews help beginners ask questions and share insights. Optional hardware labs (e.g. using a GPU or Jetson device) let participants experience deployment on real devices.