PPT s09-machine vision-s2
- 1. Course: Machine Vision Recognition (1) Session 10 D5627 – I Gede Putra Kusuma Negara, B.Eng., PhD
- 2. Outline • Pattern Recognition • Supervised Learning • Classification
- 4. What is Pattern Recognition? • A pattern is an entity, vaguely defined, that could be given a name, e.g., – Fingerprint image, – Handwritten word, – Human face – Etc. • Pattern recognition is the study of how machines can – Observe the environment, – Learn to distinguish patterns of interest, – Make sound and reasonable decisions about the categories of the patterns.
- 5. Pattern Recognition Application (example) Problem Domain Application Input Pattern Pattern Classes Document image analysis Optical character recognition Document image Characters, words Document classification Internet search Text document Semantic categories Document classification Junk mail filtering Email Junk/non-junk Multimedia database retrieval Internet search Video clip Video genres Speech recognition Telephone directory assis- tance Speech waveform Spoken words Natural language processing Information extraction Sentences Parts of speech Biometric recognition Personal identification Face, iris, fingerprint Authorized users for access control Medical Computer aided diagnosis Microscopic image Cancerous/healthy cell Military Automatic target recognition Optical or infrared image Target type Industrial automation Printed circuit board inspection Intensity or range image Defective/non-defective product Industrial automation Fruit sorting Images taken on a conveyor belt Grade of quality Remote sensing Forecasting crop yield Multispectral image Land use categories Bioinformatics Sequence analysis DNA sequence Known types of genes Data mining Searching for meaningful patterns Points in multidimensional space Compact and well- separated clusters
- 6. Example • Problem: Sorting incoming fish on a conveyor belt according to species. • Assume that we have only two kinds of fish: – Sea bass – Salmon
- 7. Decision Process • What kind of information can distinguish one species from the other? – length, width, weight, number and shape of fins, tail shape, etc. • What can cause problems during sensing? – lighting conditions, position of fish on the conveyor belt, camera noise, etc. • What are the steps in the process? – capture image → isolate fish → take measurements → make decision
- 8. Selecting Features • Assume a fisherman told us that a sea bass is generally longer than a salmon. • We can use length as a feature and decide between sea bass and salmon according to a threshold on length. • How can we choose this threshold?
- 9. Selecting Features • How can we choose the threshold l∗ to make a reliable decision? • Even though sea bass is longer than salmon on the average, there are many examples of fish where this observation does not hold.
- 10. Selecting Features • Try another feature: average lightness of the fish scales. • It looks easier to choose the threshold x∗ but we still cannot make a perfect decision
- 11. Cost of Error • We should also consider costs of different errors we make in our decisions. • For example, if the fish packing company knows that: – Customers who buy salmon will object vigorously if they see sea bass in their cans. – Customers who buy sea bass will not be unhappy if they occasionally see some expensive salmon in their cans. • How does this knowledge affect our decision?
- 12. Multiple Features • Assume we also observed that sea bass are typically wider than salmon. • We can use two features in our decision: – lightness: x1 – width: x2 • Each fish image is now represented as a point (feature vector ) in a two-dimensional feature space. X = x1 x2 æ è ç ç ö ø ÷ ÷
- 13. Multiple Features • We can draw a decision boundary to divide the feature space into two regions. Does it look better than using only lightness • Does adding more features always improve the results? – Avoid unreliable features. – Be careful about correlations with existing features. – Be careful about measurement costs – Be careful about noise in the measurements • Is there some curse for working in very high dimensions?
- 14. Decision Boundaries • Can we do better with another decision rule? • More complex models result in more complex boundaries • We may distinguish training samples perfectly but how can we predict how well we can generalize to unknown samples?
- 15. Decision Boundaries • How can we manage the tradeoff between complexity of decision rules and their performance to unknown samples? • Different criteria lead to different decision boundaries
- 18. Supervised Learning • Supervised learning is the machine learning task of inferring a function from labeled training data. The training data consist of a set of training examples. • In supervised learning, each example is a pair consisting of an input object (typically a vector) and a desired output value (also called the supervisory signal) • A supervised learning algorithm analyzes the training data and produces an inferred function, which can be used for mapping new examples • An optimal scenario will allow for the algorithm to correctly determine the class labels for unseen instances. • This requires the learning algorithm to generalize from the training data to unseen situations in a "reasonable" way
- 19. Supervised Learning Steps • Data acquisition and sensing: – Measurements of physical variables. – Important issues: bandwidth, resolution, sensitivity, distortion, SNR, latency, etc. • Pre-processing: – Removal of noise in data. – Isolation of patterns of interest from the background. – Color to grayscale conversion • Feature extraction: – Finding a new representation in terms of features
- 20. Supervised Learning Steps • Model learning and estimation: – Learning a mapping between features and pattern groups and categories • Classification: – Using features and learned models to assign a pattern to a category. • Post-processing: – Evaluation of confidence in decisions. – Exploitation of context to improve performance. – Combination of experts.
- 21. Classification
- 22. Classification Problem • Apply a prediction function to a feature representation of the image to get the desired output: f( ) = “Brick” f( ) = “Fabric” f( ) = “Grass”
- 23. Testing stage Training stage Classification Steps Real world Features Decision Model Features Training Data Data acquisition/training Pre-processing Feature extraction Classification Post-processing Model learning/estimation Feature extraction/selection Pre-processing
- 24. Texture Classification (example) • We are going to classify texture image • Image source: 40 texture images from MIT VisTex database – Number of classes: 5 (brick, fabric, grass, sand, stone) – Train set: 30 images (6 images from each class) – Test set: 10 images (2 images from each class) • Features: – Gray-level co-occurrence matrix (GLCM) – Fourier power spectrum (#ring: 4, #sector:8) • Classification method: – K-nearest neighbor – Naïve-bayes • Evaluation: classification accuracy
- 25. Design Cycle Data collection: • Collecting training and testing data. • How can we know when we have adequately large and representative set of samples? Collect data Select features Select Model Train Classifier Evaluate classifier
- 26. Testing images Training images Data Collection Brick Fabric Grass Sand Stone
- 27. Design Cycle Feature selection: • Domain dependence and prior information. • Computational cost and feasibility. • Discriminative features: – Similar values for similar patterns. – Different values for different patterns. • Invariant features with respect to translation, rotation and scale. • Robust features with respect to occlusion, distortion, deformation, and variations in environment. Collect data Select features Select Model Train Classifier Evaluate classifier
- 28. Feature Selection Gray-level co-occurrence matrix (GLCM) Fourier power spectrum Features Formula Energy Inertia Inverse diff moment Entropy Pij 2 å Pij å *(i- j)2 Pij å / 1+(i- j)2 { } Pij 2 *(-lnPij ) å #ring = 4 #sector = 8
- 29. Design Cycle Model selection: • Domain dependence and prior information. • Definition of design criteria. • Parametric vs. non-parametric models. • Handling of missing features. • Computational complexity. • Types of models: templates, decision-theoretic or statistical, syntactic or structural, neural, and hybrid. • How can we know how close we are to the true model underlying the patterns? Collect data Select features Select Model Train Classifier Evaluate classifier
- 30. Model Selection k-Nearest neighbor • Assign label of nearest training data point to each test data point Naïve bayes (algorithm) • For each value yk – Estimate P(Y = yk) from the data. – For each value xij of each attribute Xi • Estimate P(Xi=xij | Y = yk) • Classify a new point via: • In practice, the independence assumption doesn’t often hold true, but Naïve Bayes performs very well despite it. x x x x x x x x o o o o o o o x2 x1 + + 1-Nearest neighbor 3-Nearest neighbor i k i k y new y Y X P y Y P Y k ) | ( ) ( max arg
- 31. Design Cycle Training: • How can we learn the rule from data? • Supervised learning: a teacher provides a category label or cost for each pattern in the training set. • Unsupervised learning: the system forms clusters or natural groupings of the input patterns. • Reinforcement learning: no desired category is given but the teacher provides feedback to the system such as the decision is right or wrong. Collect data Select features Select Model Train Classifier Evaluate classifier
- 32. Design Cycle Evaluation: • How can we estimate the performance with training samples? • How can we predict the performance with future data? • Problems of over-fitting and generalization. Collect data Select features Select Model Train Classifier Evaluate classifier
- 33. Evaluation Classification Method K-Nearest neighbor Naïve Bayes Feature Accuracy F score Accuracy F score GLCM Fourier Power Spectrum
- 34. Acknowledgment Some of slides in this PowerPoint presentation are adaptation from various slides, many thanks to: 1. Selim Aksoy, Department of Computer Engineering, Bilkent University (http://www.cs.bilkent.edu.tr/~saksoy/) 2. James Hays, Computer Science Department, Brown University, (http://cs.brown.edu/~hays/)
- 35. Thank You