CSCI-B 657 Computer Vision
3 credits
- Prerequisite(s): None
- Delivery: On-Campus
- Equivalent(s): CSCI 55700 Image Processing and Computer Vision
Description
Concepts and methods of machine vision as a branch of artificial intelligence. Basics of digital image processing. Local and global tools for deriving information from image data. Model-based object recognition and scene understanding.
Topics
Introduction to computer vision
- What is computer vision?
- A brief history of computer vision
- Book overview and syllabus
Image formation
- Geometric primitives and transformations
- Photometric image formation
- The digital camera
Image processing
- Point operators and linear filtering
- Nonlinear filtering and Fourier transforms
- Pyramids and wavelets
- Geometric transformations
Model fitting and optimization
- Scattered data interpolation
- Variational methods and regularization
- Markov random fields
Deep learning
- Supervised learning and unsupervised learning
- Deep neural networks
- Convolutional networks and more complex models
Recognition
- Instance recognition and image classification
- Object detection and semantic segmentation
- Video understanding and vision and language
Feature detection and matching
- Points and patches, edges, and contours
- Contour tracking and lines, vanishing points
- Segmentation
Image alignment and stitching
- Pairwise alignment and image stitching
- Global alignment and compositing
Motion estimation
- Translational alignment and parametric motion
- Optical flow and layered motion
Computational photography
- Photometric calibration
- High dynamic range imaging
- Super-resolution, denoising, and blur removal
- Image matting and compositing
- Texture analysis and synthesis
Structure from motion and SLAM
- Geometric intrinsic calibration and pose estimation
- Two-frame structure from motion
- Multi-frame structure from motion
- Simultaneous localization and mapping
Depth Estimation
- Epipolar geometry
- Sparse correspondence and dense correspondence
- Local methods and global optimization
- Deep neural networks and multi-view stereo
- Monocular depth estimation
3D reconstruction
- Shape from X and 3D scanning
- Surface representations and point-based representations
- Volumetric representations and model-based reconstruction
- Recovering texture maps and albedos
Image-based rendering
- View interpolation and layered depth images
- Light fields and lumigraphs
- Environment mattes and video-based rendering
- Neural rendering
Review and future directions
- Recap of key concepts
- Discussion of future directions in computer vision
Learning Outcomes
- Compare the strengths and limitations of different image processing techniques, assessing their suitability for solving specific computer vision challenges and justifying your choices based on theoretical insights. CS 4
- Investigate and interpret the underlying theories of feature extraction and matching algorithms, conducting in-depth analyses of their performance in various scenarios and making informed recommendations for optimizations. CS 1
- Appraise computer vision technologies’ ethical and societal implications, considering issues related to bias, privacy, and security, and propose informed strategies to mitigate these concerns in design and deployment. CS 6
- Conduct a comprehensive comparative evaluation of multiple 3D reconstruction methods, considering accuracy, scalability, and computational efficiency, and propose novel approaches to enhance the reliability of 3D scene reconstruction. CS 4
- Create a robust and efficient real-time object tracking system by designing innovative algorithms that fuse multiple sensory inputs, demonstrating an ability to model and predict object dynamics in complex environments. CS 1
- Develop a customized end-to-end computer vision pipeline for a specific application domain, leveraging deep learning frameworks and designing architecture variations to achieve superior performance and adaptability. CS 2
Policies and Procedures
Please be aware of the following linked policies and procedures. Note that in individual courses instructors will have stipulations specific to their course.