Download as PDF, PPTX
![26 Image representation Images are stored as a 2D array of pixels on computers. Each pixel has a certain value representing its intensity. Example of grayscale representation: • Image is black and white with shades of gray in between. • Pixel intensity is a number between 0 (black) and 255 (white). int[ ][ ] array = { {255, 170, 170, 0}, {220, 80, 80, 170}, {255, 80, 0, 0}, {175, 20, 170, 0} }; © Copyright IBM Corporation 2019, 2021](https://image.slidesharecdn.com/ieeeeed2021aiincomputervisionsamehzaghloul-ibm-211027212621/75/IEEE-EED2021-AI-use-cases-in-Computer-Vision-26-2048.jpg)
Uploaded bySAMeh Zaghloul
IEEE EED2021 AI use cases in Computer Vision
The document presents an overview of computer vision, including its history, tasks, tools, and the deep learning approaches utilized for image recognition such as convolutional neural networks (CNNs). It discusses various applications like object detection, image retrieval, and optical character recognition, along with the computer vision pipeline steps from image acquisition to classification. Additionally, it highlights tools and resources for image annotation and pretrained models for computer vision tasks.
IEEE EED2021 AI use cases in Computer Vision
- 1. 1 AI uses cases in ComputerVision Presented By: SAMeh Zaghloul – |ABCD> - IBM linkedin.com/in/sameh-zaghloul-00b5151
- 2. 2 Computer Vision: HowMachines See?
- 3. 3 Computer Vision: HowMachines See?
- 4.
- 5. 5 Deep Learning inHieroglyph Recognition
- 6. 6 Hieroglyph Alphabet Recognition– Early Stages “Ibn Wahshiyya” 985 CE “Arabic” translation of the Ancient Egyptian Hieroglyph alphabet
- 7.
- 8.
- 9. 9 Hieroglyph Alphabet Recognition– Raw Data
- 10. 10 Hieroglyph Alphabet Recognition– Training Data
- 11. 11 Hieroglyph Alphabet -Sample Tagged/Labeled Datasets
- 12. 12 Hieroglyph Alphabet –Sample Tagged/Labeled Dataset
- 13. 13 Deep Learning inHieroglyph Recognition
- 14.
- 15. 15 Computer vision history Startedwith the emergence of AI In 1956. Convolutional neural networks were proposed in the well-known 1998 research paper by Yann LeCun and Léon Bottou. - 99.2% recognition accuracy on the MNIST data set. © Copyright IBM Corporation 2019, 2021
- 16. 16 Computer vision tasks Objectdetection and recognition: Detect certain patterns within the image. Examples: - Detecting red eyes when taking photos in certain conditions. - Face recognition. © Copyright IBM Corporation 2019, 2021
- 17. 17 Computer vision tasks(cont.) Content-based image retrieval: Image retrieval from a database based on user’s image query. - By using image actual feature contents such as colors, shapes, and textures - Not using image metadata (keywords, tags, or descriptions) Optical character recognition (OCR): Converting hand- written text to a digital format. © Copyright IBM Corporation 2019, 2021
- 18. 18 Computer vision tasks(cont.) Object tracking: Following the position changes of a target object from one frame to another in an image sequence or video. The following photo shows an example of human tracking. © Copyright IBM Corporation 2019, 2021
- 19. 19 Computer vision tasks Imagerestoration: Fixing and restoring images that are corrupted by noise, such as motion blur, to their default state. Scene reconstruction: Creation of a 3D model by supplying the system with multiple 2D images from different views. The computer constructs a 3D model based on those images. (Sinha, Steedly, Szeliskiet al. 2008) © 2008 ACM. © Copyright IBM Corporation 2019, 2021
- 20. 20 Computer Vision tools OpenCV:CV open source library - C++, Python, Java, and MATLAB interfaces PyTorchCV is based on PyTorch framework. • Used for various computer vision tasks. • Includes a collection of pretrained models for image classification, segmentation, detection, and pose estimation. scikit-image is an open source library for image processing. • Includes a set of algorithms for image processing. • Implements algorithms and utilities that are used in research, education, and industry applications. • Well-documented API in the Python programming language. © Copyright IBM Corporation 2019, 2021
- 21. 21 Computer Vision –Deep Learning – CNN (Convolutional Neural Network) © Copyright IBM Corporation 2019, 2021 Convolution is a Sliding-Window over a Matrix-Representation of an Image Loose Mimicking of the Overlapping Tiling of the Biological Visual Field
- 22. 22 Computer Vision AnnotationTools IBM Cloud Annotations cloud.annotations.ai - Makes labeling images and training machine learning models easy. - Supports both photos and videos. - Object detection: An object detection model provides the following information: • Type of object and prediction confidence level • Location: The coordinates and area of where the object is in the image. • Count: The number of objects found in the image. • Size: How large the object is with respect to the image dimensions. - Object classification: An object classification model provides the following information • Type of object and prediction confidence level © Copyright IBM Corporation 2019, 2021
- 23. 23 Computer Vision AnnotationTools VoTT - Visual Object Tagging Tool https://vott.z22.web.core.windows.net/#/ CVAT - Computer Vision Annotation Tool https://cvat.org/ Google’s Teachable Machine https://teachablemachine.withgoogle.com/ © Copyright IBM Corporation 2019, 2021
- 24. 24 Computer Vision TrainedModels Public Datasets Model Zoo https://models.roboflow.com/ Computer Vision Datasets https://public.roboflow.com/ Image Datasets for Computer Vision Training https://imerit.net/blog/22-free-image-datasets-for-computer-vision-all-pbm/ © Copyright IBM Corporation 2019, 2021
- 25.
- 26. 26 Image representation Images arestored as a 2D array of pixels on computers. Each pixel has a certain value representing its intensity. Example of grayscale representation: • Image is black and white with shades of gray in between. • Pixel intensity is a number between 0 (black) and 255 (white). int[ ][ ] array = { {255, 170, 170, 0}, {220, 80, 80, 170}, {255, 80, 0, 0}, {175, 20, 170, 0} }; © Copyright IBM Corporation 2019, 2021
- 27. 27 Image representation (cont.) -Example of color representation: • The pixel color is represented as a mix of Red, Green, and Blue. • The pixel intensity becomes three 2D arrays or one 2D array, where each entry is an object containing the 3 color values of RGB © Copyright IBM Corporation 2019, 2021
- 28. 28 Computer vision pipeline Thesteps and functions that are included are highly dependent on the application. Here is a conventional visual pattern recognition pipeline. Image Acquisition Pre- Processing Segmentation Feature Extraction & Selection Classification © Copyright IBM Corporation 2019, 2021
- 29. 29 Computer vision pipeline(cont.) 1. Image acquisition: The process of acquiring images and saving them in a digital image format for processing. Images often use common formats, such as .jpeg, .png, and .bmp. Images are 2D images that are stored as arrays of pixels according to their color model. © Copyright IBM Corporation 2019, 2021
- 30. 30 Computer vision pipeline(cont.) 2. Pre-processing: Preparing the image for the processing stage Examples: - Resizing images - Noise reduction - Contrast adjustment © Copyright IBM Corporation 2019, 2021
- 31. 31 Computer vision pipeline(cont.) 3. Segmentation: Partitioning an image into regions of similarity. Grouping pixels and features with similar characteristics together. Helps with selecting regions of interest within the images. These regions can contain objects of interest that we want to capture. Segmenting an image into foreground and background. © Copyright IBM Corporation 2019, 2021
- 32. 32 Computer vision pipeline(cont.) 4. Feature extraction and selection: Find distinguishing information about the image. Image features examples: distinct color in an image or a specific shape such as a line, edge, corner, or an image segment. © Copyright IBM Corporation 2019, 2021
- 33. 33 Computer vision pipeline(cont.) 5. Classification: The extracted features are used to classify the image. More processing might be done on the classified images to identify more features from the image. Example: After face detection, identify features on the face, such as hair style, age, and gender. © Copyright IBM Corporation 2019, 2021
- 34.
- 35.
- 36.
- 37.
- 38.