Free online courseModern Computer Vision and Deep Learning for Images - CNNs, RNNs, 3D Vision, Detection

Course Description

Modern Computer Vision and Deep Learning for Images is a free online course by NPTEL in Technology and Programming, designed for learners who want a solid, end-to-end understanding of how machines interpret images. It bridges classical computer vision foundations with modern deep learning methods, helping you build intuition for both the math and the practical design choices behind today’s vision systems.

You will progress from neural network essentials such as neurons, multilayer perceptrons, regression, training workflows, gradient descent, activation functions, backpropagation, optimization strategies, and regularization techniques including dropout and preprocessing. From there, the course dives into convolutional neural networks, covering core CNN concepts, properties, and influential architectures, then expands into sequence modeling with recurrent networks, encoder-decoder approaches, and LSTMs to connect deep learning ideas across different data structures.

On the classical vision side, it explores low-level image processing topics like spatial and frequency domain filtering, edge and line detection, and feature detection and description with well-known methods such as Harris corners, blob detection, SIFT, and SURF. The geometry and 3D vision portion introduces camera modeling and intrinsics, 2D geometric transformations, two-view stereo, epipolar geometry, fundamental matrix estimation, and structure from motion, including factorization methods, bundle adjustment, and dense 3D reconstruction. Finally, it moves into mid-level vision and modern deep approaches for optical flow, segmentation, clustering with GMMs, and object detection, giving you a broad toolkit for real-world computer vision problems.