Deep Density-based Image Clustering

11 Dec 2018  ·  Yazhou Ren, Ni Wang, Mingxia Li, Zenglin Xu ·

Recently, deep clustering, which is able to perform feature learning that favors clustering tasks via deep neural networks, has achieved remarkable performance in image clustering applications. However, the existing deep clustering algorithms generally need the number of clusters in advance, which is usually unknown in real-world tasks. In addition, the initial cluster centers in the learned feature space are generated by $k$-means. This only works well on spherical clusters and probably leads to unstable clustering results. In this paper, we propose a two-stage deep density-based image clustering (DDC) framework to address these issues. The first stage is to train a deep convolutional autoencoder (CAE) to extract low-dimensional feature representations from high-dimensional image data, and then apply t-SNE to further reduce the data to a 2-dimensional space favoring density-based clustering algorithms. The second stage is to apply the developed density-based clustering technique on the 2-dimensional embedded data to automatically recognize an appropriate number of clusters with arbitrary shapes. Concretely, a number of local clusters are generated to capture the local structures of clusters, and then are merged via their density relationship to form the final clustering result. Experiments demonstrate that the proposed DDC achieves comparable or even better clustering performance than state-of-the-art deep clustering methods, even though the number of clusters is not given.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Clustering Fashion-MNIST DDC-DA Accuracy 0.609 # 11
NMI 0.661 # 7
Image Clustering Fashion-MNIST DDC Accuracy 0.619 # 9
NMI 0.682 # 6
Image Clustering LetterA-J DDC-DA Accuracy 0.691 # 1
NMI 0.629 # 1
Image Clustering LetterA-J DDC Accuracy 0.573 # 2
NMI 0.546 # 2
Image Clustering MNIST-full DDC-DA NMI 0.961 # 5
Accuracy 0.986 # 5
Image Clustering MNIST-test DDC NMI 0.916 # 5
Accuracy 0.965 # 5
Image Clustering MNIST-test DDC-DA NMI 0.927 # 3
Accuracy 0.97 # 3
Image Clustering USPS DDC NMI 0.918 # 8
Accuracy 0.967 # 7
Image Clustering USPS DDC-DA NMI 0.939 # 4
Accuracy 0.977 # 4

Methods