BOK.md

• Lect 1: Introduction to Computer Vision

  • what is computer vision?
    • human vision
    • what is it related to?
  • why is computer vision difficult?
  • why do we need to study CV?
  • course overview
    • Introduction
    • Panorama stitching
      • Local interset point detectors
      • Local feature descriptors and matching
    • Monocular camera calibration
      • Projective geometry
      • Nonlinear least squares
      • Measurement using a single camera
    • Binocular stereo vision
    • Machine Learning
      • Basic for machine learning and its applications
      • Applications of DCNNs
  • Introduction to numerical geometry

• Lect 2: Local Interest Point Detectors

  • Local Invariant Features
  • Motivation
    • Requirement
    • Invariance
  • Harris Corner Detector
    • Finding corners
    • Basic idea
    • Mathematics
    • Algorithm
    • Steps
    • Properties
    • Local descriptors for Harris corners
  • Scale Invariant Point Detection
    • Automatic scale selection
    • Laplacian-of-Gaussian detector
    • Difference-of-Gaussian detector

• Lect 3: Local Feature Descriptors and Matching

  • Scale Invariant Feature Transform(SIFT)
    • Assign Keypoints Orientations
    • Building the descriptor
    • Application of SIFT
  • Case Study: Homography Estimation
    • Matrix Differentiation
    • Lagrange Multiplier
      • LS for Inhomogeneous Linear System
      • LS for Homogeneous Linear System
    • Least-squares for Linear Systems
    • RANSAC-based Homography Estimation

• Lect 4: Math Prerequisite 1 - Projective Geometry

  • Vector Operations
  • Fundamentals of Projective Geometry

• Lect 5: Math Prerequisite 2 - Nonlinear Least-squares

  • Non-linear Least Squares
    • General Methods for Non-linear Optimization
      • Basic Concepts
      • Descent Methods
        • 2-phase methods
          • steepest descent to compute the descent direction
          • Newton's method to compute the descent direction
          • Line search to find the step length
        • 1-phase methods
          • trust region method
          • damped method
    • Non-Linear Least Squares Problems
      • Basic Concepts
      • Gauss-Newton Method
      • Levenberg-Marquardt Method
      • Powell's Dog Leg Method

• Lect 6: Measurement Using a Single Camera

  • What is Camera Calibration?
  • Modeling for Image Pipeline
  • The General Framework for Camera Calibration
  • Initial Rough Estimation of Calibration Parameters
  • Nonlinear Least-squares
  • Bird's-eye-view Generation

• Lect 7: Basic for Machine Learning and A Special Emphasis on CNN

  • Basic concepts
    • A little history about AI
    • What is machine learning?
    • SUpervised vs Unsupervised
    • Training, testing, and validation
    • Overfitting, generalization, and capacity
    • Performance evaluation
    • Class-imbalance issue
  • Linear model
    • Linear regression
    • Logistic regression
    • Softmax regression
      • Cross Entropy
  • Neural network
  • Convolutional neural network(CNN)
  • Modern CNN architectures
  • DCNN for object detection
    • Background
    • R-CNN
    • Faster-RCNN
    • Yolo

• Lect 8: Applications of CNNs

  • Vision-based Parking-slot Detection
    • Background Introduction
    • General Flowchart
    • Surround-view Synthesis
    • Parking-slot Detection from Surround-view
    • Experiments
  • Human-body Keypoint Detection
    • Problem definition
    • OpenPose

• Lect 9: Introduction to Numerical Geometry

  • Introduction
  • Basic concepts in geometry
  • Discrete geometry
    • Metric for discrete geometry
    • Sampling
  • Rigid shape analysis
    • Euclidean isometries removal
    • ICP-based shape matching