---

100 Colab's for Computer Vision

---

100 Google colab's for image processing, pattern recognition and computer vision

by Domingo Mery, Gabriel Garib, Christian Pieringer, Sebastian Pulgar, Javier Tramon

Table of Contents

Image Processing

• Basic Image Processing
• Spatial Domain
• Frequency Domain
• Image Restoration
• Morphology
• Segmentation
• Color Processing

Pattern Recognition

• Feature Extraction
• Feature Selection and Transformation
• Classification
• Performance Evaluation
• Clustering

Geometry in Computer Vision

• Basics
• Geometric Transformations
• One View
• Two Views
• Multiple Views

Deep Learning in Computer Vision

• Image Classification
• Transfer Learning
• Facial Analysis
• Human Analysis
• Object Detection
• Tracking
• Segmentation
• GAN
• Anomaly Detection

---

Image Processing

+ Basic Image Processing

• How to read color images (+)

• Introductory Example: Rice segmentation (+)

• Sampling (spatial and grayscale sampling) (+)

• Basic color segmentation (+)

+ Image Processing in Spatial Domain

• Geometric Transformation (+)

• Equalization of Histogram (+)

• Filtering with Masks (kernels) (+)

+ Image Processing in Frequency Domain

• 1D Convolution (+)

• 1D Fourier Transform for Audio Signals (+)

• 2D Filtering uisng Fourier (+)

• Homomorphic Filter (+)

+ Image Restoration

• Basic Image Restoration (+)

• Wiener Filter (+)

+ Morphology

• Erosion, Dilation, Opening, Closing, Skeletization, Filling-Holes, Gradient (+)

• Top Hat Filter (+)

• Median Filter (+)

+ Segmentation

• Region Growing (+)

• Region Detection using OTSU, MSER, etc. (+)

• See Segmentation in Module [ Deep Learning > Segmentation ]

• Edge Detection (+)

• Hough Transform (+)

• Watershed (+)

• Motion Segmentation (+)

+ Color Processing

• Color Segmentation using K-means (+)

• Color Segmentation using High Contrast Images (+)

• Color Image Enhancement (+)

(+) Tested in July, 2023

---

Pattern Recognition

+ Feature Extraction

• Geometric features (basic, elliptical and moments) (+)

• Fourier Descriptors (+)

• Intensity Features (basic, contrast and Crossing Line Profile) (+)

• Face Recognition using Local Binary Patterns (LBP) (+)

• Texture Recognition using LBP, Haralick and Gabor features (+)

• Pedestrian Detection using Histogram of Gradients (HoG) (+)

• Cow Biometrics using SIFT (+)

+ Feature Selection and Transformation

• 5-Character Recognition using Sequential Forward Selection (SFS) (+)

• Basic Feature Selection and Transformation: SFS, Exhaustive Search, PCA, PLSR, ICA, etc. (+)

• Face Recognition with LBP using PCA, ICA and PLSR (+)

+ Classification

• Collection of Methods for Visualization of Feature Space (+)

• Basic Classifiers (KNN, Bayes, LDA, QDA, Mahalanobis) (+)

• Setting Hyperparameter K of KNN using Validation Dataset (+)

• Neural Networks from scratch (+)

• Neural Networks (using sklearn) (+)

• Support Vector Machines (SVM) (+)

• Exhaustive Search of the Best Classifier (+)

• Neural Networks for MNIST dataset (+)

• See Bag of Visual Words [ Pattern Recognition > Clustering ]

+ Performance Evaluation

• Accuracy Estimation (Hold-Out, Cross-Val, Leave-one-out) and Model Selector (+)

• Model Selector (+)

• Data Augmentation (+)

+ Clustering

• Collection of Clustering Algorithms (+)

• Bag of Visual Words (+)

• See Face Clustering in Module [ Deep Learning > Facial Analysis ]

(+) Tested in July 2023

---

Geometry in Commputer Vision

+ Basics

• Points and lines in homogeneous coordinates (*)

• Pose orientation of a face using Landmarks (*)

• Parameter estimation with and without outliers (RANSAC) (*)

+ Geometric Transformations

• 2D Transformations and Homography (*)

• 3D Transformations (*)

• 3D Reconstruction (*)

+ One View

• Perspective Correction using Homography (*)

• Simple Camera Calibration (*)

• Pro Calibration (*)

+ Two Views

• Epipolar Geometry (two views) (*)

• Estimation of Fundamental Matrix using Corresponding Points in Two Views (*)

+ Multiple Views

• Mosaics using SIFT (*)

• See Epipolar Geometry and Fundamental Matrix [ Multiple View Geometry > Two Views ]

• Trifocal Geometry (three views) (*)

(*) Updated in July-2023, some of them in Spanish

---

Deep learning in Computer Vision

+ Image Classification

• Classification of Eyes and Noses (two classes) (+)

• Defect detection in aluminum wheels (two classes) (+)

• Skin Lesion Recognition (two classes) (+)

• Classification of Dogs and Cats (two classes) (+)

• Covid recognition in Lung X-ray images (three classes) (+)

• Classification of Pedestrians with Bikes (three classes) (+)

• Skin Lesion Recognition (seven classes) (+)

• ARLNET (Attention Residual Learning blocks) (+)

• EfficientNet (+)

• Visual Transformers (from library HuggingFaces) (+)

• Classification examples using pre-trained mmodels (ResNet*, VGG*, ShuffleNet*, etc.) (+)

(+) Tested in July, 2023

+ Facial Analysis

• Face Detection using MTCNN (+)

• Face Recognition (1:1) with ArcFace (face verification) (*)

• Face Recognition (1:1) with AdaFace (face verification) (+)

• Face Recognition (1:N) with AdaFace (face recognition) (+)

• Searching a Face in a Gallery with AdaFace (+)

• Face Clustering (+)

• Age and Gender Recognition (*)

• Face Expression Recognition (+)

• Landmark Detection (68 face landmarks) (+)

• Face Geometric Mesh (+)

• Estimation of Head Pose from Face Landmarks (+)

• Mask Detection in Face using YOLOv5 (+)

• Eye and Mouth Detection using YOLOv5 (+)

• Restoration of Image Faces using GFPGAN (+)

• Face Analysis Explanation (Minus, Plus, AVG, SEQ, LIME, RISE) (+)

(+) Tested in July, 2023 (*) There are some version troubles :(

+ Human Analysis

• Detection of Human Pose using OpenPose

• See Classification of Pedestrians with Bikes in Module [ Deep Learning > Image Classification ]

+ Object detection

• Detection of Eye and Mouth using YOLOv5

• Defect Detection in Aluminum Castings using YOLOv5

• Detection of Threat Objects in Baggage using YOLOv5

• General Object Detection using YOLOv5 (without training)

• OCR - Optical Chracter Recognition (OCR) using pytesseract

+ Tracking

• Tracking of Multiple Objects in Videos using YOLOv5

+ Segmentation (using Deep Learning)

• Segmentation of Skin Lesions using UNet

+ Generative adversarial network (GAN)

• Basic GAN for Digits Generation using MNIST

• DCGAN for Digits Generation using MNIST

• SN-GAN for Digits Generation using MNIST

• WGAN-GP for Digits Generation using MNIST

• DCGAN for Generation of X-ray images of Shuriken (64x64 pixeles)

+ Anomaly Detection

• Anomaly Detection using MNIST(0: normal class, 1,2,..9: anomaly)

• Contrastive Learning in CIFAR dataset