Download as PDF, PPTX
![IMAGE COVARIANCE MATRIX • Optimization Problem: Maximize the trace of covariance matrix (Sx) ( ) { [( )( ) ]}T xtr tr E E E S Y Y Y Y ( ) { [( )( ) ]} { [( ) ( ) ]} { [ ( ) ( ) ]} { [( ) ( )] } { } T x T T T T T T T tr tr E E E tr E E E tr E E E tr E E E tr S Y Y Y Y A A XX A A X A A A A X X A A A A X X GX Y = AX ( ) ( )tr XY tr YX 1 1 ( ) ( ) M T k k kM G A A A A](https://image.slidesharecdn.com/ml-in-ip-160701164115/75/Machine-learning-in-image-processing-25-2048.jpg)
Uploaded byData Science Thailand
Machine learning in image processing
The document discusses machine learning applications in image processing, including feature extraction and various machine learning approaches for tasks like face recognition and object detection. It highlights tools such as MATLAB and OpenCV along with techniques like SIFT and SURF for feature detection. Key topics include supervised vs unsupervised learning, classification, and regression techniques relevant to image data.
In this document
Powered by AI
Overview of the presentation on Machine Learning in Image Processing, covering topics like feature extraction, approaches, and applications like face recognition and object detection.
Defines key concepts: Classification and Regression. Discusses Supervised vs Unsupervised learning, with detailed structure of data and target classification.
Discusses methods for feature extraction including SIFT, eigenvectors, and various features like color, texture, and shape, along with alignment considerations.
Lists classification algorithms such as K-NN, Neural Networks, SVM, and CNN, highlighting approaches for machine learning in image processing.
Explains two main approaches: Image to Non-Image and Image to Image, detailing tasks like object detection and image enhancement.
Explores applications including Face Recognition and Hallucination, Object Detection, and Augmented Reality, including methods used like Viola Jones for detection.
Introduces various tools supporting machine learning in image processing, including OpenCV, MATLAB, and alternatives like Python with OCR capabilities.
Machine learning in image processing
- 1. MACHINE LEARNING I N IMAGE PROCESSING PAR I N YA S A N G U A N S AT
- 2. Asst. Parinya Sanguansat,Ph.D. Computer Engineering, Panyapiwat Institute of Management
- 3.
- 4. CONTENTS • Introduction • FeatureExtraction • Machine Learning approaches – Image to image – Image to non-image • Applications – Face Recognition – Face Hallucination – Object Detection – Augmented Reality • Tools
- 5. INTRODUCTION • Classification • Regression MachineDataClass Label MachineData Data Discrete Continues
- 6.
- 7. INTRODUCTION • Supervised Learning •Unsupervised Learning MachineTraining Data learning Training Target learningTest Data classify Test Target MachineTraining Data learningTest Data cluster Data Cluster
- 8.
- 9. FEATURE EXTRACTION • Normaldata • Image data A1 A2 A3 A4 A5 A6 O1 1 2 1 1 2 3 O2 1 4 2 5 3 1 O3 2 1 5 2 1 3 O4 3 2 4 5 2 4 O1 O2 O3
- 10. VECTORIZATION O1 O2 O3 O1 O2 O3 PROBLEMS: •High-Dimensional feature vector • Very large memory • Very long processing time • Singularity problem • Small Sample Size problem
- 11. SCALE INVARIANT FEATURE TRANSFORM(SIFT) • To detect and describe local features in an images, wildly used in image search, object recognition, video tracking, gesture recognition, etc. • Speeded Up Robust Features (SURF)
- 12.
- 13. BAG OF (VISUAL)WORDS
- 14. OTHER FEATURE EXTRACTIONS • Color •Texture • Shape • Statistic
- 15.
- 16. CLASSIFIERS • K-NN • Neuralnetwork • SVM • CNN
- 17.
- 18. IMAGE TO NON-IMAGE MachineLearningImage Information Object detection and tracking Image recognition and classification
- 19. IMAGE TO IMAGE MachineLearningImage Image Image retrieval Image enhancement Extrapolated art (http://extrapolated-art.com/)
- 20.
- 21.
- 22. FACE RECOGNITION PreprocessingFace imageFeature Extraction Classifier Label
- 23. PCA Crop & Resize m n Vectorize mn M M 1 1 ( )() M T k k kM C Covariance matrix Dimension = mn x mn ( ), 1,2,3, ,T i i d y = x max( )d MPC Scalar
- 24. 2DPCA Crop & Resize m n VectorizeM 1 1 ( )( ) M T k k kM G A A A A Image covariance matrix Dimension = n x n , 1,2,3, ,i i i dY = Ax max( )d nPCV Vector
- 25. IMAGE COVARIANCE MATRIX •Optimization Problem: Maximize the trace of covariance matrix (Sx) ( ) { [( )( ) ]}T xtr tr E E E S Y Y Y Y ( ) { [( )( ) ]} { [( ) ( ) ]} { [ ( ) ( ) ]} { [( ) ( )] } { } T x T T T T T T T tr tr E E E tr E E E tr E E E tr E E E tr S Y Y Y Y A A XX A A X A A A A X X A A A A X X GX Y = AX ( ) ( )tr XY tr YX 1 1 ( ) ( ) M T k k kM G A A A A
- 26.
- 27.
- 28. CCTV SAMPLES • Asianon Asian database • European on Asian database
- 29. OBJECT DETECTION • ViolaJones method Positive samples Negative samples Cascade Classifier (Adaboost)
- 30. AUGMENTED REALITY matchFeatures estimateGeometricTransform +imwarp detectSURFFeatures Create Marker
- 31.
- 32.
- 33.
- 34. ALTERNATIVE TO MATLAB •Opensource Mostly compatible Different Syntax PythonBrowser-based
- 35.