An evaluation of large-scale methods for image instance and class discovery
This work addresses the challenge of unsupervised image grouping for instance and class discovery, offering incremental improvements in clustering methods for large-scale datasets.
The paper tackles the problem of discovering meaningful subsets of related images from large, unannotated collections, evaluating methods like Markov Clustering against k-means, and shows that Markov Clustering consistently outperforms others, with efficient GPU implementation reducing discovery costs even for 100 million images.
This paper aims at discovering meaningful subsets of related images from large image collections without annotations. We search groups of images related at different levels of semantic, i.e., either instances or visual classes. While k-means is usually considered as the gold standard for this task, we evaluate and show the interest of diffusion methods that have been neglected by the state of the art, such as the Markov Clustering algorithm. We report results on the ImageNet and the Paris500k instance dataset, both enlarged with images from YFCC100M. We evaluate our methods with a labelling cost that reflects how much effort a human would require to correct the generated clusters. Our analysis highlights several properties. First, when powered with an efficient GPU implementation, the cost of the discovery process is small compared to computing the image descriptors, even for collections as large as 100 million images. Second, we show that descriptions selected for instance search improve the discovery of object classes. Third, the Markov Clustering technique consistently outperforms other methods; to our knowledge it has never been considered in this large scale scenario.