Innovative Techniques for Image Clustering and Classification
Muhammad Akram, Sibghat Ullah Bazai, Samina Samina, Wajid Hassan Moosa · Advances in computational intelligence and robotics book series · 2025
The chapter is a review of techniques in deep leaning for tasks such as classification and clustering. Basically, due to the discussion of the two main topics in deep learning, the chapter is divided into two parts, one discussing the clustering methods such as first a basic understanding of clustering method is made then moving towards autoencoder based architectures that includes variational autoencoders (VAE), k-means with autoencoders, then self-organizing maps, spectral clustering and DBSCAN. The other part of the chapter is focused on classification methods, where the architecture of a convolutional neural network (CNN) is discussed, proceeding to ResNet, DenseNet and EfficientNet, a little touch of transformer-based CNN is discussed, a part of these vision transformers and capsule networks are mentioned. A comparison of both the methods, i.e., clustering and classification is discussed, that will make it clearer that how both these methods are different from one another.