This page outlines the weekly schedule for lectures, labs, assignments, and examinations. The schedule will be updated regularly to align with the University of Juba's academic calendar and holiday schedule. Reading materials, lecture slides, and lab materials will be accessible through this schedule, with links provided for downloading prior to the commencement of each lecture or lab session. If you encounter any difficulties or have questions, please contact the lead Teaching Fellow, Thiong Abraham.
This lecture introduces object detection, a computer vision task that extends beyond image classification by not only identifying the objects present in an image but also locating their spatial extent through bounding boxes. Building upon the foundational concepts of image classification, where a model learns to assign a single label to an entire image, object detection models learn to simultaneously classify and localize multiple objects. We'll explore techniques like sliding windows, region proposal networks, and modern architectures such as YOLO and Faster R-CNN, detailing how these methods utilize convolutional neural networks to extract features, predict object classes, and refine bounding box coordinates. The lecture will also cover evaluation metrics like mean Average Precision (mAP) and practical applications of object detection, from autonomous driving to industrial automation.
This lecture introduces Vision Transformers (ViTs), a cutting-edge approach to image classification that leverages the power of transformer architectures, originally developed for natural language processing, to analyze visual data. The session will cover the fundamental principles of Vision Transformers, including their architecture, self-attention mechanisms, and how they differ from traditional convolutional neural networks (CNNs). We will explore the advantages of ViTs, such as their ability to capture global context and their scalability to large datasets, as well as their challenges, including computational requirements and data efficiency.
This lecture provides a comprehensive introduction to image classification, a core computer vision task, beginning with an overview of its fundamental principles and diverse applications. The session delves into the algorithmic foundations, highlighting the pivotal role of Convolutional Neural Networks (CNNs) in learning hierarchical image features. Various CNN architectures, including ResNet, VGG, and Inception, are explored, emphasizing their unique strengths and design considerations. The lecture culminates in a hands-on lab exercise where participants apply their newfound knowledge to classify African wildlife images, utilizing a provided dataset and pre-trained models. This practical component allows students to solidify their understanding of CNNs and their application in real-world scenarios, specifically focusing on the identification of species like buffalo, elephant, rhino, and zebra, thereby bridging theoretical concepts with practical implementation.
This module introduces the course materials to students and gives an overview of Computer Vision, its tasks, applications, and the state of the art.