Free online courseConvolutional Neural Networks for Computer Vision: Detection, Recognition

Course Description

Convolutional Neural Networks for Computer Vision: Detection, Recognition is a free online course in Technology and Programming, focused on Artificial Intelligence and Machine Learning. It is designed to help you understand how CNNs power modern computer vision systems, from core image operations to practical model design choices used in real-world applications.

You will build intuition for fundamental convolution concepts, including edge detection, padding, strided convolutions, and working with multi-channel volumes. The course explains how a convolutional network layer operates, how pooling helps with representation, and why convolutions are especially effective for visual data. Along the way, you will connect these building blocks to end-to-end CNN examples to see how the pieces fit together.

The course then shifts toward influential architectures and proven engineering patterns in deep vision. You will explore classic networks as well as modern approaches like residual connections, network-in-network ideas, and inception-style designs. Practical strategies such as using open-source implementations, transfer learning, and data augmentation are introduced to help you train stronger models with less data and effort, while keeping an eye on broader trends shaping the current state of computer vision.

For detection tasks, you will learn the key ideas behind object localization and landmark detection, then move into object detection workflows. Topics include sliding-window style convolutional implementations, Intersection over Union, non-max suppression, anchor boxes, and the intuition behind popular methods such as YOLO and region proposal approaches.

The final portion introduces face recognition concepts, including one-shot learning, Siamese networks, triplet loss, and face verification. You will also be introduced to neural style transfer, breaking down content and style cost functions and considering generalizations beyond standard 2D images. By the end, you will have a strong conceptual map of CNN-based vision systems and the tools to confidently approach recognition and detection problems.