CS 195: Computer Vision
Fall 2024
Instructor: Alimoor Reza
Assistant Professor of Computer Science
Department of Mathematics and Computer Science
Drake University

Class room: Collier-Scripps # 335
Meeting time: Mon/Wed (3:30 pm - 4:45 pm)
Office hours: Tue/Thu (12:00pm-2:30pm) or by appointment


Schedule
A tentative schedule below (subject to change as we progress).


Date Topics Readings Items due
Week 1 (Mon: 08/26)

  Introduction to Computer Vision
  Lecture Slides 1
 

  Course logistics
  Introduction to computer vision

 
Week 1 (Wed: 08/28)

  Image Filtering: cross-correlation
  Lecture Slides 2
  In-class activity #1 (image filtering cross-correlation)
 

  Pillow concepts
  Pillow tutorial
  Image.open(path)
  Image.save(path)
  Image.new('L', (width, height), color=0)
  Image.getpixel( (x,y) )
  Image.putpixel( (x,y), value)
  ImageDraw( img )
 
 
Week 2 (Wed: 09/02)

  Labor Day (university holiday; classes do not meet)
 
 
 
Week 2 (Wed: 09/04)

  Image Filtering: convolution
  Lecture Slides 3
  Activity (image filtering: convolution)

  Notebook#1 (released on 09/05)
 
  Image.histogram()

 
Week 3 (Mon: 09/09)

  Image Recognition
  Image Classification and Neural Networks (NN)
  Types of NN: Multilayer Perceptron (MLP)
  Lecture Slides 4
  In-class activity #2 (PyTorch MLP)
 
  PyTorch Basics (review this before attempting activity#2)
  PyTorch: nn.Linear()
  PyTorch: nn.Sigmoid()
  PyTorch: nn.Tanh()
  PyTorch: nn.ReLU()
 
Week 3 (Wed: 09/11)

  In-class activity #2 (PyTorch MLP) continued
  Types of NN: Convolution Neural Network (CNN)
  Lecture Slides 5
 

  PyTorch: nn.Conv2d()
  PyTorch: nn.MaxPool2d()
  PyTorch: nn.AvgPool2d()
 

  Notebook#1 (due on 09/13)
Week 4 (Mon: 09/16)

  Reza is traveling (class does not meet)
  Midwest Computer Vision Workshop@Indiana University Bloomington

  Notebook #2 (released on 09/16)
 

 

 
Week 4 (Wed: 09/18)

  In-class activity #3 (PyTorch CNN)
  Notebook#2 discussion

 
Week 5 (Mon: 09/23)

  Popular Image Classification Networks (AlexNet, VGG, ResNet)
  Dissecting Popular Pre-trained CNN models
  Lecture Slides 8
  AlexNet dissection
  VGG-16 dissection
  ResNet dissection


 
  Notebook#2 (due on 09/23)
Week 5 (Wed: 09/25)

  Fine-tuning a Pre-trained CNN model
  Lecture Slides 9
  Fine-tuning CNNs

  Notebook#3 (Fine-tuning popular CNNs)


 
Week 6 (Mon: 09/30)

  Vision Transformer (ViT)
  Other ViT variants (Swin Transformer Tiny)
  ConvNext, Inflated-ConvNext Tiny
  Lecture Slides 10
  Vision Transformer (ViT) Dissection
 
  The Hardware Lottery - Sara Hooker
  Transformer (designed to win the hardware lottery)
  Vision Transformer (ViT): Transformers for image recognition at scale
 
 
Week 6 (Wed: 10/02)

  Object Detection: Classical and Deep Learing-Based
  Lecture Slides 11a: Classical Detectors
  Lecture Slides 11b: Deep learning-based Detectors (Faster R-CNN, Mask R-CNN)
 

 
  Notebook#3 (due on 10/02)
 
Week 7 (Mon: 10/07)
 
  Inference with Popular Object Detectors (Faster R-CNN, Mask R-CNN)
  Lecture Slides 12
  Inference with Object Detector: Faster R-CNN

 


Week 7 (Wed: 10/09)

  Lecture Slides 13
  Fine-tuning Popular Object Detectors (Faster R-CNN, Mask R-CNN)
  Fine-tuning Object Detector: Faster R-CNN

  Quiz#1 (released on 10/09)


 
Week 8 (Mon: 10/14)

  Fall Break (no class)
 
 
 
Week 8 (Wed: 10/16)
 

  Image Segmentation
  Classical Segmentation (Felzenswalb, SLIC, clustering)
  Lecture Slides 13 part1
  Lecture Slides 13 part2
  Classical Image Segmentation Codes



 
  Quiz#1 (due on 10/16)
Week 9 (Mon: 10/21)

  Vision Foundation Model for Image Segmentation: Segment Anything (SAM)
  Lecture Slides 14 [pptx]


 
Week 9 (Wed: 10/23)

  Inference with SAM: Automatically generating object masks with Segment Anything (SAM)
  Semantic Segmentaiton
  Popular Semantic Segmentation Models (FCN, SegNet, U-Net, PSPNet)
  Lecture Slides 15



 
Week 10 (Mon: 10/28)

  Probability Basics
  Discrete Probability Distribution
  Lecture Slides 16 part1
  Discrte probability distribution code
  Continous Probability Distribution
  Gaussian Distributions
  Lecture Slides 16 part2
  In-class activity#5: Continous probability distribution
 
  Notebook#4 (Fine-tuning Detectors: Faster & Mask R-CNN)


 
Week 10 (Wed: 10/30)

  Classes do not meet in-person
  Dr. Reza is traveling to China as part of a Drake University delegation



 
Week 11 (Mon: 11/04)

  Classes do not meet in-person
  Dr. Reza is traveling to China as part of a Drake University delegation

 
Week 11 (Wed: 11/06)

  Image Generation and Deep Generative Models: Generative Adversarial Network (GAN)
  Lecture Slides 17

 
 
Week 12 (Mon: 11/11)

  Maximum Likelihood (ML) Estimate
  Lecture Slides 18 part1
  In-class activity#6: Maximum likelihood estimate for Gaussian distribution

  Image Generation and Deep Generative Models: Variational Auto Encoder (VAE)
  Lecture Slides 18 part2


  Tutorial on variational autoencoder by Carl Doersch
  Generating Counterfactual Images: C2C-VAE - CBREIS'22
  Generative model: implicit reparameterization
  Wasserstein Distance
 
Week 12 (Wed: 11/13)

  Image Generation and Deep Generative Models: Variational Auto Encoder (VAE)
  Lecture Slides 18 part2 (continued)
  In-class activity#7: VAE code for digit image generation using MNIST dataset

  Quiz#2 (released on 11/13)

 
 
  Notebook#4 (due on 11/13)
Week 13 (Mon: 11/18)

  Image Generation and Deep Generative Models: Diffusion Model
  Lecture Slidees 19
  In-class activity#8: Generating images from text prompt using Stable Diffusion Model
 
 
  Diffusion models are evolutionary algorithms - arxiv'24

 
  Quiz#2 (due on 11/19)
Week 13 (Wed: 11/20)

  Image Transformation
  Linear transformation
  Lecture Slidees 20
  In-class activity#8: Rotation, Scaling, Shearing transformation

 
 
Week 14 (Mon: 11/25)

  Notebook #5 (Semantic Segmentation Inference and Training)

  From Images to 3D Models-ACM Communication'2002

 
Week 14 (Wed: 11/27)

  Thanksgiving Break (no class)

 
 
Week 15 (Mon: 12/02)

  Homogenous coordinates
  Affine and projective transformation
  Lecture Slides 21 part1
 


  Camera Projection
  Modeling projection
  Activity: projection practice
  Generalizing the model
  Adding a lens
  Lecture Slides 21 part2
 

 
 
Week 15 (Wed: 12/04)

  3D Reconstruction from Images: Stereo Matching
  Depth from stereo
  The case of parallel cameras
  Parallel cameras and rectification

  Naive stereo matching
  Stereo matching with deep neural network
  Lecture Slides

 
  Notebook#5 (due on 12/04)
  Final Project Presentation on 12/11
Week 16 (Mon: 12/09)

  Quiz#3 (released on 12/10)
  No final exam!



Week 16 (Wed: 12/11)

  Final project group presentation

  Course Evaluation

  @Collier-Scripps#335


  Q3 (due on 12/13)
  Final Project Code +
  Presentation Slides (due by 12/16)