AD-701 AI for Computer Vision - Dec 2024 Question Paper

Under each question, How to answer lists the points an answer should cover, outlined by CampusPrep from the syllabus. Check the details against your textbook.

1. 7 Marks
   aDefine Computer Vision and explain its significance in modern technology.
   How to answer

   • Definition of computer vision as enabling machines to interpret visual data
   • Core tasks: image classification, detection, segmentation, tracking
   • Significance with modern examples: healthcare, autonomous vehicles, surveillance
   • Role of AI/deep learning in advancing capabilities
2. 7 Marks
   bDiscuss future trends in AI for computer vision and their potential impact.
   How to answer

   • Current state of AI for vision: CNNs, transformers, foundation models
   • Future trends: self-supervised learning, multimodal, edge AI, 3D vision
   • Potential impact on industry, healthcare, robotics
   • Challenges: ethics, bias, compute
3. 7 Marks
   aHow can image-based rendering be applied in virtual reality?
   How to answer

   • Definition of image-based rendering (IBR)
   • Use of captured images to synthesize novel views for VR
   • Steps: capture, reconstruction, rendering
   • Benefits for realism and immersion in VR
4. 7 Marks
   bDescribe the concept of linear filtering in image processing.
   How to answer

   • Concept of linear filtering as convolution with kernel
   • Formula for 2D convolution/correlation
   • Examples: mean, Gaussian, Sobel as linear filters
   • Properties: linearity, shift-invariance
5. 7 Marks
   aDefine pyramids and wavelets and their role in image processing.
   How to answer

   • Definition of image pyramids: Gaussian, Laplacian
   • Definition of wavelets and multi-resolution analysis
   • Role in compression, denoising, feature extraction
   • Advantage of scale-space representation
6. 7 Marks
   bWhat are geometric transformations? List common types.
   How to answer

   • Definition of geometric transformations preserving geometry
   • Common types: translation, rotation, scaling, affine, perspective
   • Mathematical representation using transformation matrices
   • Applications in registration, augmentation
7. 7 Marks
   aIllustrate the process of edge detection using neighborhood operators.
   How to answer

   • Concept of edges and intensity discontinuities
   • Neighborhood operators: Sobel, Prewitt, Roberts, Laplacian
   • Steps: smoothing, gradient computation, non-max suppression
   • Illustrative example of mask operation
8. 7 Marks
   bExplain how active contours are used in image segmentation.
   How to answer

   • Definition of active contours/snakes
   • Energy minimization: internal and external forces
   • Evolution to object boundaries
   • Use in segmentation of medical and tracking images
9. 7 Marks
   aWhat is the mean shift algorithm and how is it applied in computer vision?
   How to answer

   • Definition of mean shift as non-parametric mode seeking
   • Iterative shift to density maximum using kernel
   • Application in segmentation, tracking, clustering
   • Advantages: no prior cluster number
10. 7 Marks
    bExplain the concept of normalized cuts and its advantages.
    How to answer

    • Concept of graph-based segmentation with normalized cut criterion
    • Formula minimizing cut cost normalized by association
    • Advantages over min-cut: avoids small isolated segments
    • Applications in perceptual grouping
11. 7 Marks
    aHow do energy-based methods contribute to feature detection?
    How to answer

    • Concept of energy functional for feature saliency
    • Examples: snake energy, MRF energy, Harris response
    • Optimization to locate corners, edges, blobs
    • Robustness to noise and illumination
12. 7 Marks
    bCompare and contrast view interpolation and layered depth images.
    How to answer

    • Definition of view interpolation for novel views
    • Definition of layered depth images with multiple depth layers
    • Comparison on basis of rendering quality, storage, occlusion handling
    • Use cases in IBR and VR
13. 7 Marks
    aExplain the process of video-based rendering and its significance.
    How to answer

    • Definition of video-based rendering using video streams
    • Pipeline: capture, depth estimation, view synthesis
    • Significance for dynamic scenes, telepresence, VR
    • Challenges of temporal consistency
14. 7 Marks
    bAnalyze the techniques for category recognition and their applications.
    How to answer

    • Categories: template matching, bag-of-words, CNNs, transformers
    • Analysis of accuracy, scalability, generalization
    • Applications in object detection, retrieval, robotics
    • Datasets and evaluation metrics
15. 7 Marks
    aHow do recognition databases aid in improving computer vision algorithms?
    How to answer

    • Role of large annotated datasets: ImageNet, COCO
    • Training, benchmarking and generalization improvement
    • Enabling deep learning and transfer learning
    • Evaluation and bias reduction
16. 7 Marks
    bCompare triangulation methods and their applications in 3D reconstruction.
    How to answer

    • Principle of triangulation from multiple views
    • Methods: linear, optimal, robust triangulation
    • Comparison on accuracy, noise sensitivity, speed
    • Applications in SfM, stereo, 3D reconstruction