Computer Vision: Object Detection In Adversarial Vision

Cover
Half Title
Title Page
Copyright Page
Dedications
Table of Contents
Preface
Acknowledgements
Targeted Readership
About the Author
Chapter 1: Fundamentals of Object Detection
	1.1 Defining Computer Vision and Object Detection
	1.2 Objects and Approaches of Object Detection
	1.3 Representation of Objects
		1.3.1 Point-Based Representation of Object
		1.3.2 Rectangular Bounding Box–Based Representation of Object
		1.3.3 Elliptical Shape–Based Representation of Object
		1.3.4 Contour-Based Representation of Object
		1.3.5 Mask-Based Representation of Object
		1.3.6 Skeleton-Based Representation of Object
	1.4 Applications of Object Detection
		1.4.1 Autonomous Vehicles
		1.4.2 Surveillance and Security
		1.4.3 Retail and Inventory Management
		1.4.4 Medical Imaging
		1.4.5 Industrial Automation
		1.4.6 Augmented Reality
		1.4.7 Robotics
		1.4.8 Agriculture
		1.4.9 Human–Computer Interaction
		1.4.10 Wildlife Monitoring and Conservation
	1.5 Challenges of Object Detection
		1.5.1 Quality/Accuracy-Based Challenges
		1.5.2 Efficiency-Based Challenges
	1.6 Organization of the Book
	References
Chapter 2: Background of Degradation
	2.1 Defining Degradation
	2.2 Categorization of Degradation
		2.2.1 Noise
			2.2.1.1 Distribution
			2.2.1.2 Correlation
			2.2.1.3 Nature
			2.2.1.4 Source
		2.2.2 Blur
		2.2.3 Distortions
	2.3 Mechanism of Degradation
		2.3.1 Gaussian Noise
		2.3.2 Rayleigh Noise
		2.3.3 Erlang (or Gamma) Noise
		2.3.4 Exponential Noise
		2.3.5 Uniform Noise
		2.3.6 Impulse (Salt and Pepper) Noise
	2.4 Effect of Image Degradation in Object Detection
		2.4.1 Target Object Information Loss in Object Detection Task
		2.4.2 Inaccurate Localization of Far Distant and Small Objects in Detection Task
		2.4.3 Atmospheric Turbulence in Object Detection Task
	Homework Problems
	References
Chapter 3: Imaging Modalities for Object Detection
	3.1 Visual Imaging Modality
	3.2 Infrared Imaging Modality
	3.3 CCTV Surveillance Imaging Modality
	3.4 Unmanned Aerial Vehicle (UAV) Imaging Modality
	References
Chapter 4: Real-Time Benchmark Datasets for Object Detection
	4.1 Indoor Datasets and Their Key Characteristics
		4.1.1 VSSN 2006 (Video Surveillance & Sensor Networks 2006) Dataset
		4.1.2 CAVIAR (Context Aware Vision Using Image-based Active Recognition) Dataset
		4.1.3 i-LIDS (Imagery Library for Intelligent Detection Systems) Dataset
		4.1.4 SBM-RGBD Dataset
		4.1.5 ADE20K (ADE20K-Scene Parsing) Dataset
		4.1.6 RGB-D Scene Understanding (Sun RGB-D) Dataset
		4.1.7 NYU Depth V2 Dataset
		4.1.8 Stanford 2D-3D-Semantics Dataset
	4.2 Outdoor Datasets and Their Key Characteristics
		4.2.1 CD.Net 2014 (Change Detection. Net) Dataset
		4.2.2 BMC 2012 (Background Models Challenge 2012) Dataset
		4.2.3 PETS 2009 (Performance Evaluation of Tracking and Surveillance 2009) Dataset
		4.2.4 I2R (Institute for Infocom Research) Dataset
		4.2.5 ETISEO (Evaluation of the Treatment and Interpretation of Video Sequences) Dataset
		4.2.6 DAVIS (Densely Annotated Video Segmentation) Dataset
		4.2.7 Wallflower Dataset
		4.2.8 ViSal (Video-based Saliency) Dataset
		4.2.9 SegTrack (Segments Track) Dataset
		4.2.10 SegTrack V2 (Segments Track Version 2) Dataset
		4.2.11 FBMS (Freiburg-Berkley Motion Segmentation) Dataset
		4.2.12 VOS (Video-based Salient Object Detection) Dataset
		4.2.13 Fish4Knowledge Dataset
		4.2.14 ViSOR (Video Surveillance Online Repository) Dataset
		4.2.15 BEHAVE Dataset
		4.2.16 MarDCT (Maritime Detection, Classification and Tracking) Dataset
		4.2.17 LASIESTA (Labeled and Annotated Sequences for Integral Evaluation of Segmentation Algorithms) Dataset
		4.2.18 REMOTE SCENE IR Dataset
		4.2.19 CAMO-UOW Dataset
		4.2.20 Grayscale-Thermal Foreground Detection (GTFD) Dataset
		4.2.21 Extended Tripura University Video Dataset (E-TUVD)
		4.2.22 OSU-T (OSU Thermal Pedestrian) Dataset
		4.2.23 BU-TIV (Thermal Infrared Video) Dataset
		4.2.24 ASL-TID Dataset
		4.2.25 Tripura University Video Dataset at Nighttime (TU-VDN)
		4.2.26 COCO (Common Objects in Context) Dataset
		4.2.27 PASCAL VOC (Visual Object Classes) Dataset
		4.2.28 KITTI Dataset
	References
Chapter 5: Artifacts Impact on Different Object Visualization
	5.1 Background of Artifacts
	5.2 Artifacts with Respect to Object Detection in Degraded Vision
		5.2.1 Artifacts in Captured Images and Videos
			5.2.1.1 Indoor Environment
			5.2.1.2 Outdoor Environment
	5.3 Impact of Different Artifacts in Objects Visualization
		5.3.1 Poor Illumination/Lighting
		5.3.2 Weather Condition
		5.3.3 Poor Illumination
		5.3.4 Camera Jitter
		5.3.5 Motion Blur
		5.3.6 Object Overlapping or Occlusion
		5.3.7 Camouflage Effect
		5.3.8 Small Object Identification
		5.3.9 Deformation
		5.3.10 Background Clutter
	Homework Problems
	References
Chapter 6: Visibility Enhancement of Images in Degraded Vision
	6.1 Fundamental of Visibility Restoration
	6.2 Background of Visibility Restoration
	6.3 Multiple Image Approaches for Visibility Enhancement
		6.3.1 Diverse Climate Condition (DCC)-Based Methods
			6.3.1.1 Chromatic Framework–Based DCC
			6.3.1.2 Bad Weather Vision–Based DCC
		6.3.2 Polarization-Based Methods
		6.3.3 Depth and Transmission Map-based Methods
	6.4 Single-Image Approaches for Visibility Enhancement
		6.4.1 Image Enhancement–Based Model
			6.4.1.1 Non-Model-Based Visibility Restoration
			6.4.1.2 Model-Based Visibility Enhancement
		6.4.2 Contrast Restoration–Based Model
			6.4.2.1 Dark Channel Prior (DCP)
			6.4.2.2 CLAHE (Contrast-Limited Adaptive Histogram Equalization)
		6.4.3 Deep Learning–Based Model
			6.4.3.1 DehazeNet
			6.4.3.2 Multiscale Deep CNN (MSCNN)
			6.4.3.3 Gated Fusion Network (GFN)
			6.4.3.4 Model-Driven Deep Visibility Restoration Approach
			6.4.3.5 Dehazing Using Reinforcement Learning System (DDRL)
			6.4.3.6 Haze Concentration Adaptive Network (HCAN)
	6.5 Performance Evaluation Metrics
		6.5.1 Full Reference Matrix
			6.5.1.1 Peak Signal to Noise Ratio (PSNR)
			6.5.1.2 Mean Square Error (MSE)
			6.5.1.3 Structural Similarity Index (SSI)
			6.5.1.4 Mean Absolute Error (MAE)
		6.5.2 No-Reference Matrix
			6.5.2.1 Saturated Pixel Percentage (ρ)
			6.5.2.2 Perceptual Haze Density
			6.5.2.3 Contrast Gain (CG)
			6.5.2.4 Visible Edges Ratio
			6.5.2.5 Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE)
			6.5.2.6 Natural Image Quality Evaluator (NIQE)
			6.5.2.7 Perception-Based Image Quality Evaluator (PIQUE)
	Homework Problems
	References
Chapter 7: Object Detection in Degraded Vision
	7.1 Background Modeling–Based Approaches for Object Detection
		7.1.1 Classical Methods Based on Background Modeling for Object Detection
			7.1.1.1 Background Subtraction
			7.1.1.2 Frame Differencing
		7.1.2 Deep Learning Methods Based on Background Modeling for Object Detection
	7.2 Location-Oriented or Bounding Box–Based Approaches to Object Detection
		7.2.1 Classical Methods Based on Location-Oriented or Bounding Box–Based Approaches to Object Detection
			7.2.1.1 Kalman Filter
			7.2.1.2 Particle Filter
			7.2.1.3 Optical Flow
		7.2.2 Deep Learning Methods Based on Location-Oriented or Bounding Box–Based Approaches for Object Detection
			7.2.2.1 Region-Based Convolutional Neural Network (R-CNN)
			7.2.2.2 Fast Region-Based Convolutional Neural Network (Fast R-CNN)
			7.2.2.3 Faster Region-Based Convolutional Neural Network (Faster R-CNN)
			7.2.2.4 Mask RCNN
			7.2.2.5 Single-Shot MultiBox Detector (SSD)
			7.2.2.6 You Only Look Once (YOLO)
			7.2.2.7 Adaptive Weighted Residual Dilated Network (AWRDNet)
	7.3 Performance Evaluation Measures for Object Detection
	7.4 Performance Comparison of Published Results of State-of-the-Art Methods for Object Detection
	Homework Problems
	References
Chapter 8: Hands-on Practical for Object Detection Approaches in Degraded Vision
	8.1 Deep Learning Algorithms
		8.1.1 Convolution Neural Network for Binary/Multi-Class Classification Problem
		8.1.2 Deep Learning Architectures Used for Binary/Multi-Class Classification
			8.1.2.1 Visual Geometry Group-16 (VGG-16)
			8.1.2.2 Visual Geometry Group-19 (VGG-19)
			8.1.2.3 Residual Network-50 (ResNet-50)
		8.1.3 Deep Learning Architectures Used for Object Detection
			8.1.3.1 Region-Based Convolutional Neural Network (R-CNN)
			8.1.3.2 Fast Region-Based Convolutional Neural Network (Fast R-CNN)
			8.1.3.3 Faster Region-Based Convolutional Neural Network (Faster R-CNN)
			8.1.3.4 Mask Region-Based Convolutional Neural Network (Mask R-CNN)
			8.1.3.5 Single-Shot Multibox Detector (SSD)
			8.1.3.6 You Only Look Once (YOLO)
	8.2 Practical Approaches and Applications
		8.2.1 Introduction to Python
		8.2.2 Installation of Python
		8.2.3 Source Codes of Deep Learning–Based Algorithms for Classification and Object Detection
			8.2.3.1 Source Code of Basic CNN
			8.2.3.2 Source Code of VGG-16
			8.2.3.3 Source Code of VGG-19
			8.2.3.4 Source Code of ResNet-50
			8.2.3.5 Source Code of R-CNN
			8.2.3.6 Source Code of Fast R-CNN
			8.2.3.7 Source Code of Faster R-CNN
			8.2.3.8 Source Code of Mask R-CNN
			8.2.3.9 Source Code of SSD
			8.2.3.10 Source Code of YOLO-V1
	References
Index