Segmentation of VHR EO Images using Unsupervised Learning

TL;DR

This paper introduces an unsupervised semantic segmentation approach for Earth observation images that requires only a single unlabeled scene, leveraging deep clustering and contrastive learning to produce accurate segmentation maps without labeled data.

Contribution

The novel method enables unsupervised segmentation of large remote sensing scenes using a lightweight deep model trained on a single unlabeled image, eliminating the need for extensive labeled datasets.

Findings

Effective segmentation on Vaihingen dataset

Requires only one unlabeled scene for training

Outperforms some existing unsupervised methods

Abstract

Semantic segmentation is a crucial step in many Earth observation tasks. Large quantity of pixel-level annotation is required to train deep networks for semantic segmentation. Earth observation techniques are applied to varieties of applications and since classes vary widely depending on the applications, therefore, domain knowledge is often required to label Earth observation images, impeding availability of labeled training data in many Earth observation applications. To tackle these challenges, in this paper we propose an unsupervised semantic segmentation method that can be trained using just a single unlabeled scene. Remote sensing scenes are generally large. The proposed method exploits this property to sample smaller patches from the larger scene and uses deep clustering and contrastive learning to refine the weights of a lightweight deep model composed of a series of the convolution layers along with an embedded channel attention. After unsupervised training on the target image/scene, the model automatically segregates the major classes present in the scene and produces the segmentation map. Experimental results on the Vaihingen dataset demonstrate the efficacy of the proposed method.