Machine Learning for Graphics and Computer Vision Advanced Learning Path - MAI4CAREU Master in AI Created byLaia Güell Paule|Updated16 August 2024This Machine Learning for Graphics and Vision Learning path offers a carefully crafted curriculum that merges the realms of computer vision and computer graphics, key areas in artificial intelligence that are transforming technology and creative industries. It is designed for the Master's program in Artificial Intelligence of the University of Cyprus, which was developed with co-funding from the MAI4CAREU European project. The course provides an in-depth exploration of both theoretical and practical aspects. Over 13 weeks, this course covers essential topics such as basic regression, deep learning for image and video analysis, feature extraction, semantic understanding, and creative applications like style transfer. Additionally, it delves into the intersection of vision and graphics, exploring 3D data processing, generative networks, motion capture, and neural rendering. The curriculum is structured to progressively build students' expertise, starting with foundational concepts, and advancing to sophisticated techniques, preparing them to tackle real-world challenges and innovate in the field of graphics and vision.Part I: Introduction1.Overview of machine learning and introduction to basic regression techniquesSupervised, unsupervised, and reinforcement learningUnderstanding linearity and non-linearity in machine learning modelsPart II: Computer Vision2. Fundamentals of Computer VisionImage formation and representationBasic image processing techniquesFeature detection and matching3. Machine Learning for Image Recognition Convolutional Neural Networks (CNNs)Transfer learning and fine-tuningObject detection and segmentation4. Deep Learning for Computer VisionDeep learning techniques for image classificationadvanced CNN architectures: ResNet, Inception, DenseNetObject detection algorithms such as YOLO and Faster R-CNN5. Deep Learning for VideosDeep learning approaches for video classification6. Semantic UnderstandingDeep learning for semantic segmentationVisualize and interpret neural network layers and activationsGenerative Adversarial Networks (GANs) and their applicationsImage inpainting and saliency detection using GANsAutoencoders and their use in image denoising and generationPart III: Computer Graphics7. Machine Learning in Computer GraphicsGraphics pipelines and rendering techniques.Compositional image generation techniques.Style transfer and neural texture synthesis.8. 3D Computer Vision 3D reconstruction and depth estimation.Point clouds and 3D mesh processing.Processing irregular data structures.Applications of deep learning in 3D vision.9. Character AnimationMotion capture techniques, pose representation, and character animation.Human Pose Estimation and Activity RecognitionKeypoint detection and trackingSkeleton-based action recognitionApplications in sports, healthcare, and entertainmentDeep motion analysis and synthesisDeep reinforcement learning for animation control and physics-based animation.Part IV: Advanced Topics in Graphics and Vision10. Advanced TopicsNeural style transferTexture synthesis.Neural rendering techniques to create realistic images.Image and video super-resolutionCreative ApplicationsGenerative networks for creating faces, landscapes, and sketches.Denoising techniques in image processing.Adversarial training and open research problems. Introductory learning materialsIntermediateAdvancedMAI4CAREU - Machine Learning for Graphics and Computer Vision - Overview of Machine Learning and Introduction to Basic Regression TechniquesArtificial IntelligenceMachine LearningShow lessIn this lecture, we provide an overview of machine learning and introduce basic regression techniques. Students will explore the three primary paradigms of machine learning: supervised, unsupervised, and reinforcement learning. The module will also cover the fundamental concepts of linearity and non-linearity in machine learning models, equipping students with the knowledge to differentiate and apply appropriate regression techniques in various scenarios. This foundation is crucial for understanding more complex topics in graphics and vision. IntermediateAdvanced MAI4CAREU - Machine Learning for Graphics and Computer Vision - Fundamentals of Computer VisionArtificial IntelligenceMachine LearningShow lessIn this lecture, we delve into the fundamentals of computer vision. The session begins with an exploration of image formation and representation, providing students with an understanding of how images are captured and structured. We then cover basic image processing techniques, including filtering, edge detection, and image enhancement. The module concludes with an introduction to feature detection and matching, essential for tasks such as object recognition and image stitching. These foundational concepts are critical for advancing in the field of computer vision.IntermediateAdvancedDigital ExpertMAI4CAREU - Machine Learning for Graphics and Computer Vision - Machine Learning for Image RecognitionArtificial IntelligenceMachine LearningShow lessIn this lecture, we focus on machine learning techniques for image recognition. The module introduces Convolutional Neural Networks (CNNs), highlighting their architecture and application in image classification tasks. Students will learn about transfer learning and fine-tuning, powerful methods to leverage pre-trained models for new tasks. The session also covers advanced topics in object detection and segmentation, equipping students with the skills to develop robust models for identifying and segmenting objects within images. These techniques are fundamental for many practical applications in graphics and vision.IntermediateAdvancedDigital ExpertMAI4CAREU - Machine Learning for Graphics and Computer Vision - Machine Learning in Computer Graphics Artificial IntelligenceMachine LearningShow lessIn this lecture, we explore the integration of machine learning in computer graphics. The session begins with an overview of graphics pipelines and rendering techniques, providing foundational knowledge for creating visual content. Students will learn about compositional image generation techniques, enabling the assembly of complex scenes from individual components. The module also delves into image synthesis, highlighting how deep learning can enhance traditional rendering processes. Additionally, we cover style transfer and neural texture synthesis, demonstrating how machine learning can be used to create visually appealing and artistically coherent images.See more materialsSee less materialsAdvanced learning materialsIntermediateAdvancedMAI4CAREU - Machine Learning for Graphics and Computer Vision - Deep Learning for Computer Vision Artificial IntelligenceMachine LearningDigital skillsShow lessIn this lecture, we explore deep learning techniques specifically tailored for computer vision. The session begins with an overview of deep learning methods for image classification, providing a solid foundation in understanding how deep networks process visual data. We then dive into advanced CNN architectures such as ResNet, Inception, and DenseNet, examining their innovations and improvements over traditional models. The module also covers state-of-the-art object detection algorithms, including YOLO and Faster R-CNN, equipping students with the knowledge to implement and optimize these powerful techniques for real-world applications in computer vision.IntermediateAdvancedMAI4CAREU - Machine Learning for Graphics and Computer Vision - Deep Learning for VideosArtificial IntelligenceShow lessIn this lecture, we delve into deep learning approaches for video classification. Students will learn about the unique challenges and techniques involved in processing video data, including the temporal dynamics that distinguish video from still images. The module covers various deep learning architectures tailored for video analysis, such as Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks, providing a comprehensive understanding of how to classify and interpret complex video content effectively.IntermediateAdvancedMAI4CAREU - Machine Learning for Graphics and Computer Vision - Semantic UnderstandingArtificial IntelligenceMachine LearningShow lessIn this lecture, we explore deep learning techniques for semantic understanding. The session covers deep learning methods for semantic segmentation, enabling students to partition images into meaningful segments. We will also focus on visualizing and interpreting neural network layers and activations to gain insights into model behavior. The module introduces Generative Adversarial Networks (GANs) and their applications, including image inpainting and saliency detection. Additionally, we will delve into autoencoders, highlighting their use in image denoising and generation. This comprehensive overview equips students with advanced skills for enhancing and interpreting visual data. IntermediateAdvancedMAI4CAREU - Machine Learning for Graphics and Computer Vision - 3D Computer VisionArtificial IntelligenceMachine LearningShow lessIn this lecture, we focus on 3D computer vision. The session begins with techniques for 3D reconstruction and depth estimation, essential for creating accurate 3D models from 2D images. Students will learn about point clouds and 3D mesh processing, gaining skills in handling and manipulating 3D data. We will also cover methods for processing irregular data structures, which are common in 3D vision tasks. The module concludes with applications of deep learning in 3D vision, showcasing how advanced models can be utilized for tasks such as 3D object recognition and scene understanding.IntermediateAdvancedMAI4CAREU - Machine Learning for Graphics and Computer Vision - Character AnimationArtificial IntelligenceMachine LearningShow lessIn this lecture, we explore character animation. The session begins with an introduction to motion capture techniques, pose representation, and the fundamentals of character animation. Students will then delve into human pose estimation and activity recognition. The session begins with keypoint detection and tracking, essential for identifying and following human joints in images and videos. Students will learn about skeleton-based action recognition, which uses these keypoints to understand and classify human actions. The module also explores various applications of these techniques in sports, healthcare, and entertainment, demonstrating their practical value in real-world scenarios.Finally, students will delve into deep motion analysis and synthesis, learning how to create realistic and dynamic character movements. The module also covers deep reinforcement learning for animation control, enabling students to develop advanced physics-based animations that respond to interactive environments. This comprehensive overview equips students with cutting-edge techniques for creating lifelike character animations. IntermediateAdvancedMAI4CAREU - Machine Learning for Graphics and Computer Vision - Advanced Topics in Graphics and VisionArtificial IntelligenceMachine LearningShow lessIn this lecture, we explore advanced topics in graphics and vision. The session begins with neural style transfer, a technique that applies the artistic style of one image to the content of another. Students will then delve into texture synthesis, learning methods to create high-quality textures for 3D models. The module covers neural rendering techniques for creating realistic images, providing insights into cutting-edge rendering advancements. Finally, we address image and video super-resolution, techniques used to enhance the resolution and quality of visual content. This advanced overview equips students with the latest tools and methodologies in graphics and vision. Finally, students will focus on creative applications of machine learning. The session begins with generative networks, exploring their use in creating faces, landscapes, and sketches, enabling students to produce novel and artistic images. We then cover denoising techniques in image processing, essential for improving image quality by removing noise and artifacts. The module also addresses adversarial training, discussing its applications and challenges, and highlights open research problems in the field. This comprehensive overview empowers students to leverage machine learning for innovative and artistic purposes.See more materialsSee less materialsLearning path DetailsDigital skill levelIntermediateAdvancedDigital technology / specialisationArtificial IntelligenceMachine Learning Share this page Log in to comment
