非常全面的计算机视觉和机器学习相关的开源项目工程。
一、特征提取FeatureExtraction：
· SIFT [1] [Demo program][SIFTLibrary] [VLFeat]
· PCA-SIFT [2] [Project]
· Affine-SIFT [3] [Project]
· SURF [4] [OpenSURF] [Matlab Wrapper]
· Affine Covariant Features [5] [Oxford project]
· MSER [6] [Oxford project] [VLFeat]
· Geometric Blur [7] [Code]
· Local Self-Similarity Descriptor [8] [Oxford implementation]
· Global and Efficient Self-Similarity [9][Code]
· Histogram of Oriented Graidents [10] [INRIAObject Localization Toolkit] [OLTtoolkit for Windows]
· GIST [11] [Project]
· Shape Context [12] [Project]
· Color Descriptor [13] [Project]
· Pyramids of Histograms of OrientedGradients [Code]
· Space-Time Interest Points (STIP) [14][Project] [Code]
· Boundary Preserving Dense Local Regions[15][Project]
· Weighted Histogram[Code]
· Histogram-based Interest PointsDetectors[Paper][Code]
· An OpenCV - C++ implementation of LocalSelf Similarity Descriptors [Project]
· Fast Sparse Representation with Prototypes[Project]
· Corner Detection [Project]
· AGAST Corner Detector: faster than FASTand even FAST-ER[Project]
· Real-time Facial Feature Detection usingConditional Regression Forests[Project]
· Global and Efficient Self-Similarity forObject Classification and Detection[code]
· WαSH: Weighted α-Shapes for LocalFeature Detection[Project]
· HOG[Project]
· Online Selection of DiscriminativeTracking Features[Project]
二、图像分割ImageSegmentation：
· Normalized Cut [1] [Matlabcode]
· Gerg Mori’ Superpixel code [2] [Matlab code]
· Efficient Graph-based Image Segmentation[3] [C++ code] [Matlab wrapper]
· Mean-Shift Image Segmentation [4] [EDISON C++ code] [Matlab wrapper]
· OWT-UCM Hierarchical Segmentation [5] [Resources]
· Turbepixels [6] [Matlab code 32bit] [Matlab code 64bit] [Updated code]
· Quick-Shift [7] [VLFeat]
· SLIC Superpixels [8] [Project]
· Segmentation by Minimum Code Length [9][Project]
· Biased Normalized Cut [10] [Project]
· Segmentation Tree [11-12] [Project]
· Entropy Rate Superpixel Segmentation[13] [Code]
· Fast Approximate Energy Minimization viaGraph Cuts[Paper][Code]
· Efficient Planar Graph Cuts withApplications in Computer Vision[Paper][Code]
· Isoperimetric Graph Partitioning forImage Segmentation[Paper][Code]
· Random Walks for Image Segmentation[Paper][Code]
· Blossom V: A new implementation of aminimum cost perfect matching algorithm[Code]
· An Experimental Comparison ofMin-Cut/Max-Flow Algorithms for Energy Minimization in Computer Vision[Paper][Code]
· Geodesic Star Convexity for InteractiveImage Segmentation[Project]
· Contour Detection and Image SegmentationResources[Project][Code]
· Biased Normalized Cuts[Project]
· Max-flow/min-cut[Project]
· Chan-Vese Segmentation using Level Set[Project]
· A Toolbox of Level Set Methods[Project]
· Re-initialization Free Level SetEvolution via Reaction Diffusion[Project]
· Improved C-V active contour model[Paper][Code]
· A Variational Multiphase Level SetApproach to Simultaneous Segmentation and Bias Correction[Paper][Code]
· Level Set Method Research by ChunmingLi[Project]
· ClassCut for Unsupervised ClassSegmentation[code]
· SEEDS: Superpixels Extracted via Energy-DrivenSampling [Project][other]
三、目标检测Object Detection：
· A simple object detector with boosting [Project]
· INRIA Object Detection and LocalizationToolkit [1] [Project]
· Discriminatively Trained Deformable PartModels [2] [Project]
· Cascade Object Detection with DeformablePart Models [3] [Project]
· Poselet [4] [Project]
· Implicit Shape Model [5] [Project]
· Viola and Jones’s Face Detection [6] [Project]
· Bayesian Modelling of Dyanmic Scenes forObject Detection[Paper][Code]
· Hand detection using multiple proposals[Project]
· Color Constancy, Intrinsic Images, andShape Estimation[Paper][Code]
· Discriminatively trained deformable partmodels[Project]
· Gradient Response Maps for Real-TimeDetection of Texture-Less Objects: LineMOD [Project]
· Image Processing On Line[Project]
· Robust Optical Flow Estimation[Project]
· Where's Waldo: Matching People in Imagesof Crowds[Project]
· Scalable Multi-class Object Detection[Project]
· Class-Specific Hough Forests for ObjectDetection[Project]
· Deformed Lattice Detection In Real-WorldImages[Project]
· Discriminatively trained deformable partmodels[Project]
四、显著性检测SaliencyDetection：
· Itti, Koch, and Niebur’ saliencydetection [1] [Matlabcode]
· Frequency-tuned salient region detection[2] [Project]
· Saliency detection using maximumsymmetric surround [3] [Project]
· Attention via Information Maximization[4] [Matlabcode]
· Context-aware saliency detection [5] [Matlab code]
· Graph-based visual saliency [6] [Matlab code]
· Saliency detection: A spectral residualapproach. [7] [Matlab code]
· Segmenting salient objects from imagesand videos. [8] [Matlab code]
· Saliency Using Natural statistics. [9] [Matlabcode]
· Discriminant Saliency for VisualRecognition from Cluttered Scenes. [10] [Code]
· Learning to Predict Where Humans Look[11] [Project]
· Global Contrast based Salient RegionDetection [12] [Project]
· Bayesian Saliency via Low and Mid LevelCues[Project]
· Top-Down Visual Saliency via Joint CRFand Dictionary Learning[Paper][Code]
· Saliency Detection: A Spectral ResidualApproach[Code]
五、图像分类、聚类ImageClassification, Clustering
· Pyramid Match [1] [Project]
· Spatial Pyramid Matching [2] [Code]
· Locality-constrained Linear Coding [3] [Project] [Matlab code]
· Sparse Coding [4] [Project] [Matlab code]
· Texture Classification [5] [Project]
· Multiple Kernels for ImageClassification [6] [Project]
· Feature Combination [7] [Project]
· SuperParsing [Code]
· Large Scale Correlation ClusteringOptimization[Matlab code]
· Detecting and Sketching the Common[Project]
· Self-Tuning Spectral Clustering[Project][Code]
· User Assisted Separation of Reflectionsfrom a Single Image Using a Sparsity Prior[Paper][Code]
· Filters for Texture Classification[Project]
· Multiple Kernel Learning for ImageClassification[Project]
· SLIC Superpixels[Project]
六、抠图Image Matting
· A Closed Form Solution to Natural ImageMatting [Code]
· Spectral Matting [Project]
· Learning-based Matting [Code]
七、目标跟踪Object Tracking：
· A Forest of Sensors - Tracking AdaptiveBackground Mixture Models [Project]
· Object Tracking via Partial LeastSquares Analysis[Paper][Code]
· Robust Object Tracking with OnlineMultiple Instance Learning[Paper][Code]
· Online Visual Tracking with Histogramsand Articulating Blocks[Project]
· Incremental Learning for Robust VisualTracking[Project]
· Real-time Compressive Tracking[Project]
· Robust Object Tracking viaSparsity-based Collaborative Model[Project]
· Visual Tracking via Adaptive StructuralLocal Sparse Appearance Model[Project]
· Online Discriminative Object Trackingwith Local Sparse Representation[Paper][Code]
· Superpixel Tracking[Project]
· Learning Hierarchical ImageRepresentation with Sparsity, Saliency and Locality[Paper][Code]
· Online Multiple Support InstanceTracking [Paper][Code]
· Visual Tracking with Online MultipleInstance Learning[Project]
· Object detection and recognition[Project]
· Compressive Sensing Resources[Project]
· Robust Real-Time Visual Tracking usingPixel-Wise Posteriors[Project]
· Tracking-Learning-Detection[Project][OpenTLD/C++Code]
· the HandVu：vision-based hand gesture interface[Project]
· Learning Probabilistic Non-Linear LatentVariable Models for Tracking Complex Activities[Project]
八、Kinect：
· Kinect toolbox[Project]
· OpenNI[Project]
· zouxy09 Blog[Resource]
· FingerTracker 手指跟踪[code]
九、3D相关：
· 3D Reconstruction of a Moving Object[Paper] [Code]
· Shape From Shading Using LinearApproximation[Code]
· Combining Shape from Shading and StereoDepth Maps[Project][Code]
· Shape from Shading: A Survey[Paper][Code]
· A Spatio-Temporal Descriptor based on 3DGradients (HOG3D)[Project][Code]
· Multi-camera Scene Reconstruction viaGraph Cuts[Paper][Code]
· A Fast Marching Formulation ofPerspective Shape from Shading under Frontal Illumination[Paper][Code]
· Reconstruction:3D Shape, Illumination,Shading, Reflectance, Texture[Project]
· Monocular Tracking of 3D Human Motionwith a Coordinated Mixture of Factor Analyzers[Code]
· Learning 3-D Scene Structure from aSingle Still Image[Project]
十、机器学习算法：
· Matlab class for computing ApproximateNearest Nieghbor (ANN) [Matlab class providing interface toANNlibrary]
· Random Sampling[code]
· Probabilistic Latent Semantic Analysis(pLSA)[Code]
· FASTANN and FASTCLUSTER for approximatek-means (AKM)[Project]
· Fast Intersection / Additive KernelSVMs[Project]
· SVM[Code]
· Ensemble learning[Project]
· Deep Learning[Net]
· Deep Learning Methods for Vision[Project]
· Neural Network for Recognition ofHandwritten Digits[Project]
· Training a deep autoencoder or aclassifier on MNIST digits[Project]
· THE MNIST DATABASE of handwrittendigits[Project]
· Ersatz：deep neural networks in the cloud[Project]
· Deep Learning [Project]
· sparseLM : Sparse Levenberg-Marquardtnonlinear least squares in C/C++[Project]
· Weka 3: Data Mining Software in Java[Project]
· Invited talk "A Tutorial on DeepLearning" by Dr. Kai Yu (余凯)[Video]
· CNN - Convolutional neural networkclass[Matlab Tool]
· Yann LeCun's Publications[Wedsite]
· LeNet-5, convolutional neural networks[Project]
· Training a deep autoencoder or aclassifier on MNIST digits[Project]
· Deep Learning 大牛Geoffrey E. Hinton's HomePage[Website]
· Multiple Instance LogisticDiscriminant-based Metric Learning (MildML) and Logistic Discriminant-basedMetric Learning (LDML)[Code]
· Sparse coding simulation software[Project]
· Visual Recognition and Machine LearningSummer School[Software]
十一、目标、行为识别Object, ActionRecognition：
· Action Recognition by DenseTrajectories[Project][Code]
· Action Recognition Using a DistributedRepresentation of Pose and Appearance[Project]
· Recognition Using Regions[Paper][Code]
· 2D Articulated Human Pose Estimation[Project]
· Fast Human Pose Estimation UsingAppearance and Motion via Multi-Dimensional Boosting Regression[Paper][Code]
· Estimating Human Pose from OccludedImages[Paper][Code]
· Quasi-dense wide baseline matching[Project]
· ChaLearn GestureChallenge: Principal motion: PCA-based reconstruction of motionhistograms[Project]
· Real Time Head Pose Estimation withRandom Regression Forests[Project]
· 2D Action Recognition Serves 3D HumanPose Estimation[Project]
· A Hough Transform-Based Voting Frameworkfor Action Recognition[Project]
· Motion Interchange Patterns for ActionRecognition in Unconstrained Videos[Project]
· 2D articulated human pose estimationsoftware[Project]
· Learning and detecting shape models [code]
· Progressive Search Space Reduction forHuman Pose Estimation[Project]
· Learning Non-Rigid 3D Shape from 2DMotion[Project]
十二、图像处理：
· Distance Transforms of SampledFunctions[Project]
· The Computer Vision Homepage[Project]
· Efficient appearance distances betweenwindows[code]
· Image Exploration algorithm[code]
· Motion Magnification 运动放大 [Project]
· Bilateral Filtering for Gray and ColorImages 双边滤波器 [Project]
· A Fast Approximation of the BilateralFilter using a Signal Processing Approach [Project]
十三、一些实用工具：
· EGT: a Toolbox for Multiple ViewGeometry and Visual Servoing[Project] [Code]
· a development kit of matlab mexfunctions for OpenCV library[Project]
· Fast Artificial Neural Network Library[Project]
十四、人手及指尖检测与识别：
· finger-detection-and-gesture-recognition [Code]
· Hand and Finger Detection using JavaCV[Project]
· Hand and fingers detection[Code]
十五、场景解释：
· Nonparametric Scene Parsing via LabelTransfer [Project]
十六、光流Optical flow：
· High accuracy optical flow using atheory for warping [Project]
· Dense Trajectories VideoDescription [Project]
· SIFT Flow: Dense Correspondence acrossScenes and its Applications[Project]
· KLT: An Implementation of theKanade-Lucas-Tomasi Feature Tracker [Project]
· Tracking Cars Using Optical Flow[Project]
· Secrets of optical flow estimation andtheir principles[Project]
· implmentation of the Black and Anandandense optical flow method[Project]
· Optical Flow Computation[Project]
· Beyond Pixels: Exploring NewRepresentations and Applications for Motion Analysis[Project]
· A Database and Evaluation Methodologyfor Optical Flow[Project]
· optical flow relative[Project]
· Robust Optical Flow Estimation [Project]
· optical flow[Project]
十七、图像检索Image Retrieval：
· Semi-Supervised Distance Metric Learningfor Collaborative Image Retrieval [Paper][code]
十八、马尔科夫随机场Markov RandomFields：
· Markov Random Fields forSuper-Resolution [Project]
· A Comparative Study of EnergyMinimization Methods for Markov Random Fields with Smoothness-Based Priors [Project]
十九、运动检测Motion detection：
· Moving Object Extraction, Using Modelsor Analysis of Regions [Project]
· Background Subtraction: Experiments andImprovements for ViBe [Project]
· A Self-Organizing Approach to BackgroundSubtraction for Visual Surveillance Applications [Project]
· changedetection.net: A new changedetection benchmark dataset[Project]
· ViBe - a powerful technique forbackground detection and subtraction in video sequences[Project]
· Background Subtraction Program[Project]
· Motion Detection Algorithms[Project]
· Stuttgart Artificial BackgroundSubtraction Dataset[Project]
· Object Detection, Motion Estimation, andTracking[Project]
Feature Detection and Description
General Libraries:
· VLFeat – Implementation of variousfeature descriptors (including SIFT, HOG, and LBP) and covariant featuredetectors (including DoG, Hessian, Harris Laplace, Hessian Laplace, MultiscaleHessian, Multiscale Harris). Easy-to-use Matlab interface. See Modern features: Software – Slidesproviding a demonstration of VLFeat and also links to other software. Checkalso VLFeat hands-on session training
· OpenCV – Various implementations ofmodern feature detectors and descriptors (SIFT, SURF, FAST, BRIEF, ORB, FREAK,etc.)
Fast Keypoint Detectors for Real-time Applications:
· FAST – High-speed corner detectorimplementation for a wide variety of platforms
· AGAST – Even faster than the FAST cornerdetector. A multi-scale version of this method is used for the BRISK descriptor(ECCV 2010).
Binary Descriptors for Real-Time Applications:
· BRIEF – C++ codefor a fast and accurate interest point descriptor (not invariant to rotationsand scale) (ECCV 2010)
· ORB – OpenCVimplementation of the Oriented-Brief (ORB) descriptor (invariant to rotations,but not scale)
· BRISK – Efficient Binary descriptorinvariant to rotations and scale. It includes a Matlab mex interface. (ICCV2011)
· FREAK – Fasterthan BRISK (invariant to rotations and scale) (CVPR 2012)
SIFT and SURF Implementations:
· SIFT: VLFeat, OpenCV, Original code by David Lowe, GPUimplementation, OpenSIFT
· SURF: Herbert Bay’s code, OpenCV, GPU-SURF
Other Local Feature Detectors and Descriptors:
· VGG Affine Covariant features – Oxfordcode for various affine covariant feature detectors and descriptors.
· LIOP descriptor – Sourcecode for the Local Intensity order Pattern (LIOP) descriptor (ICCV 2011).
· Local Symmetry Features – Sourcecode for matching of local symmetry features under large variations inlighting, age, and rendering style (CVPR 2012).
Global Image Descriptors:
· GIST – Matlabcode for the GIST descriptor
· CENTRIST – Global visual descriptor forscene categorization and object detection (PAMI 2011)
Feature Coding and Pooling
· VGG Feature Encoding Toolkit – Sourcecode for various state-of-the-art feature encoding methods – including Standardhard encoding, Kernel codebook encoding, Locality-constrained linear encoding,and Fisher kernel encoding.
· Spatial Pyramid Matching – Sourcecode for feature pooling based on spatial pyramid matching (widely used forimage classification)
Convolutional Nets and Deep Learning
· EBLearn – C++Library for Energy-Based Learning. It includes several demos and step-by-stepinstructions to train classifiers based on convolutional neural networks.
· Torch7 – Provides a matlab-likeenvironment for state-of-the-art machine learning algorithms, including a fastimplementation of convolutional neural networks.
· Deep Learning - Variouslinks for deep learning software.
Part-Based Models
· Deformable Part-based Detector – Libraryprovided by the authors of the original paper (state-of-the-art in PASCAL VOCdetection task)
· Efficient Deformable Part-Based Detector – Branch-and-Boundimplementation for a deformable part-based detector.
· Accelerated Deformable Part Model –Efficient implementation of a method that achieves the exact same performanceof deformable part-based detectors but with significant acceleration (ECCV2012).
· Coarse-to-Fine Deformable PartModel – Fast approach for deformable object detection (CVPR 2011).
· Poselets – C++ and Matlab versions forobject detection based on poselets.
· Part-based Face Detector and PoseEstimation – Implementation of a unified approach for face detection, poseestimation, and landmark localization (CVPR 2012).
Attributes and Semantic Features
· Relative Attributes – Modified implementation ofRankSVM to train Relative Attributes (ICCV 2011).
· Object Bank – Implementation of object banksemantic features (NIPS 2010). See also ActionBank
· Classemes, Picodes, andMeta-class features – Software for extractinghigh-level image descriptors (ECCV 2010, NIPS 2011, CVPR 2012).
Large-Scale Learning
· Additive Kernels – Source code for fast additivekernel SVM classifiers (PAMI 2013).
· LIBLINEAR – Library for large-scale linearSVM classification.
· VLFeat – Implementation for Pegasos SVMand Homogeneous Kernel map.
Fast Indexing and Image Retrieval
· FLANN – Library for performing fastapproximate nearest neighbor.
· Kernelized LSH – Source code for KernelizedLocality-Sensitive Hashing (ICCV 2009).
· ITQ Binary codes – Code forgeneration of small binary codes using Iterative Quantization and otherbaselines such as Locality-Sensitive-Hashing (CVPR 2011).
· INRIA Image Retrieval –Efficient code for state-of-the-art large-scale image retrieval (CVPR 2011).
Object Detection
· See Part-based Models and Convolutional Nets above.
· Pedestrian Detection at 100fps – Veryfast and accurate pedestrian detector (CVPR 2012).
· Caltech Pedestrian DetectionBenchmark – Excellent resource for pedestrian detection, with various linksfor state-of-the-art implementations.
· OpenCV – Enhancedimplementation of Viola&Jones real-time object detector, with trainedmodels for face detection.
· Efficient Subwindow Search – Sourcecode for branch-and-bound optimization for efficient object localization (CVPR2008).
3D Recognition
· Point-Cloud Library – Libraryfor 3D image and point cloud processing.
Action Recognition
· ActionBank – Source code for actionrecognition based on the ActionBank representation (CVPR 2012).
· STIP Features – software for computingspace-time interest point descriptors
· Independent Subspace Analysis – Look forStacked ISA for Videos (CVPR 2011)
· Velocity Histories of Tracked Keypoints - C++ codefor activity recognition using the velocity histories of tracked keypoints(ICCV 2009)
Datasets
Attributes
· Animals with Attributes – 30,475images of 50 animals classes with 6 pre-extracted feature representations foreach image.
· aYahoo and aPascal –Attribute annotations for images collected from Yahoo and Pascal VOC 2008.
· FaceTracer – 15,000 faces annotated with 10attributes and fiducial points.
· PubFig – 58,797 face images of 200 peoplewith 73 attribute classifier outputs.
· LFW – 13,233face images of 5,749 people with 73 attribute classifier outputs.
· Human Attributes – 8,000 people with annotatedattributes. Check also this link for another dataset of humanattributes.
· SUN Attribute Database –Large-scale scene attribute database with a taxonomy of 102 attributes.
· ImageNet Attributes – Variety of attribute labels forthe ImageNet dataset.
· Relative attributes – Data for OSR and a subset ofPubFig datasets. Check also this link for the WhittleSearch data.
· Attribute Discovery Dataset – Imagesof shopping categories associated with textual descriptions.
Fine-grained Visual Categorization
· Caltech-UCSD Birds Dataset – Hundredsof bird categories with annotated parts and attributes.
· Stanford Dogs Dataset – 20,000images of 120 breeds of dogs from around the world.
· Oxford-IIIT Pet Dataset – 37category pet dataset with roughly 200 images for each class. Pixel level trimapsegmentation is included.
· Leeds Butterfly Dataset – 832images of 10 species of butterflies.
· Oxford Flower Dataset – Hundredsof flower categories.
Face Detection
· FDDB – UMassface detection dataset and benchmark (5,000+ faces)
· CMU/MIT –Classical face detection dataset.
Face Recognition
· Face Recognition Homepage – Largecollection of face recognition datasets.
· LFW – UMassunconstrained face recognition dataset (13,000+ face images).
· NIST Face Homepage – includesface recognition grand challenge (FRGC), vendor tests (FRVT) and others.
· CMU Multi-PIE – containsmore than 750,000 images of 337 people, with 15 different views and 19 lightingconditions.
· FERET – Classical face recognitiondataset.
· Deng Cai’s face dataset in Matlab Format – Easy touse if you want play with simple face datasets including Yale, ORL, PIE, andExtended Yale B.
· SCFace – Low-resolution face datasetcaptured from surveillance cameras.
Handwritten Digits
· MNIST – largedataset containing a training set of 60,000 examples, and a test set of 10,000examples.
Pedestrian Detection
· Caltech Pedestrian DetectionBenchmark – 10 hours of video taken from a vehicle,350K bounding boxes forabout 2.3K unique pedestrians.
· INRIA Person Dataset –Currently one of the most popular pedestrian detection datasets.
· ETH Pedestrian Dataset – Urbandataset captured from a stereo rig mounted on a stroller.
· TUD-Brussels Pedestrian Dataset – Datasetwith image pairs recorded in an crowded urban setting with an onboard camera.
· PASCAL Human Detection – One of20 categories in PASCAL VOC detection challenges.
· USC Pedestrian Dataset – Smalldataset captured from surveillance cameras.
Generic Object Recognition
· ImageNet –Currently the largest visual recognition dataset in terms of number ofcategories and images.
· Tiny Images – 80 million 32x32 low resolutionimages.
· Pascal VOC – One of the most influentialvisual recognition datasets.
· Caltech 101 / Caltech 256 – Popular image datasetscontaining 101 and 256 object categories, respectively.
· MIT LabelMe – Online annotation tool forbuilding computer vision databases.
Scene Recognition
· MIT SUN Dataset – MITscene understanding dataset.
· UIUC Fifteen Scene Categories – Datasetof 15 natural scene categories.
Feature Detection and Description
· VGG Affine Dataset – Widely used dataset formeasuring performance of feature detection and description. CheckVLBenchmarks for an evaluation framework.
Action Recognition
· Benchmarking Activity Recognition – CVPR2012 tutorial covering various datasets for action recognition.
RGBD Recognition
· RGB-D Object Dataset – Datasetcontaining 300 common household objects
Reference:
[1]: http://rogerioferis.com/VisualRecognitionAndSearch/Resources.html
