Alison's New App is now available on iOS and Android! Download Now

Understanding and Visualizing Convolutional Neural Networks (CNNs)

In this free online course, learn about the role of convolutional neural networks in processing images.

Publisher: NPTEL
This free online course explains the connection between convolution and neural networks in processing and classifying images. You will discover how a computer can perform image classification by looking for low-level features such as edges and curves and then build up to more abstract concepts through a series of convolutional layers. The course also illustrates various approaches for visualizing CNNs.
Understanding and Visualizing Convolutional Neural Networks (CNNs)
  • Duration

    5-6 Hours
  • Students

  • Accreditation


Share This Course And
Earn Money  

Become an Affiliate Member





View course modules


Convolution neural networks (CNNs) are mainly used for image classification and recognition because of their high accuracy. The course introduces the need for using CNNs in the processing of images. You will explore the role of CNNs in conquering the difficulties presented by other networks in sifting through images. Following this, the function of the backpropagation algorithm in CNNs is discussed, where you will be taught about the mathematical notations of a backpropagation algorithm. The evolution of CNN architectures over the last few decades is also highlighted. This will include the various types of architectures and the process of selecting the appropriate architecture based on the task at hand. This course explains how CNN architectures are used to extract information classes from a multiband raster image to create thematic maps.

Next, the course highlights the various ways for improving the CNN architectures based on earlier works. This will incorporate the elements of contemporary CNN models that have been created for extending the assistance beyond other networks. You will explore the various architectural implementations that led from the MobileNet architecture to state-of-the-art EfficientNet architecture, with the many improvements in between. The course explains the importance of visualizing a CNN model. This will include the various methods of envisioning and presenting the model predictions of a CNN. In addition to this, the course will outline the applications of these neural networks in processing images. You will discover how CNN, using the deep learning technique, outperforms other methods in analyzing images. The processes for interpreting and understanding the applications of CNNs in analyzing images are also explained.

Finally, you will be taught how visual explanations are extracted from deep networks through gradient-based localization. This will include the role of class activation maps in enhancing the clarifications for profound CNNs. The current methods for explaining CNNs are highlighted. You will explore the interpretable modes and game-theoretic backgrounds used for images, texts and tabular data for real-world experiments. Lastly, you will be taught about the various approaches to addressing the problem of saturating gradients. This will include solving a problem that refers to a function for which a more considerable input will not lead to a relevant increase in output. This is an exciting course that will interest those studying computer science or those interested in these topics. Why wait? Sign up today and start learning about the application of convolutional neural networks to various computer vision tasks.


Start Course Now