Semi-Supervised Learning for hyperspectral images by non parametrically predicting view assignment

Shivam Pande,Nassim Ait Ali Braham,Yi Wang,Conrad M Albrecht,Biplab Banerjee,Xiao Xiang Zhu

2023-06-19

Abstract:Hyperspectral image (HSI) classification is gaining a lot of momentum in present time because of high inherent spectral information within the images. However, these images suffer from the problem of curse of dimensionality and usually require a large number samples for tasks such as classification, especially in supervised setting. Recently, to effectively train the deep learning models with minimal labelled samples, the unlabeled samples are also being leveraged in self-supervised and semi-supervised setting. In this work, we leverage the idea of semi-supervised learning to assist the discriminative self-supervised pretraining of the models. The proposed method takes different augmented views of the unlabeled samples as input and assigns them the same pseudo-label corresponding to the labelled sample from the downstream task. We train our model on two HSI datasets, namely Houston dataset (from data fusion contest, 2013) and Pavia university dataset, and show that the proposed approach performs better than self-supervised approach and supervised training.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to use limited labeled samples and a large number of unlabeled samples to improve the performance of the model in the hyperspectral image (HSI) classification task. Specifically, the paper proposes a semi - supervised learning method. By non - parametrically predicting view assignments to assist the discriminative self - supervised pre - training process, it can effectively train deep - learning models even when there is less labeled data, overcoming the limitation of traditional supervised learning methods that require a large amount of labeled data.

The main contributions of the paper are as follows:
1. **Extension of the PAWS method**: The PAWS (Predicting Assignments With Support samples) method is applied in the field of hyperspectral image classification, and its performance improvement compared to supervised learning and self - supervised learning algorithms is demonstrated.
2. **Experimental verification**: Evaluations under different settings (including with and without data augmentation) are carried out on the Houston 2013 dataset to demonstrate the effectiveness of this method.

Through this method, researchers hope to improve the performance of the model in the hyperspectral image classification task while reducing the cost of manual labeling. This not only helps to solve the "curse of dimensionality" problem in hyperspectral image processing, but also promotes the development of remote sensing technology, especially in large - scale data processing and analysis.