Few-shot Multispectral Segmentation with Representations Generated by Reinforcement Learning
The task of multispectral image segmentation (segmentation of images with numerous channels/bands, each capturing a specific range of wavelengths of electromagnetic radiation) has been previously explored in contexts with large amounts of labeled data. However, these models tend not to generalize well to datasets of smaller size. In this paper, we propose a novel approach for improving few-shot segmentation performance on multispectral images using reinforcement learning to generate representations. These representations are generated in the form of mathematical expressions between channels and are tailored to the specific class being segmented. Our methodology involves training an agent to identify the most informative expressions, updating the dataset using these expressions, and then using the updated dataset to perform segmentation. Due to the limited length of the expressions, the model receives useful representations without any added risk of overfitting. We evaluate the effectiveness of our approach on several multispectral datasets and demonstrate its effectiveness in boosting the performance of segmentation algorithms.
updated: Mon Nov 20 2023 15:04:16 GMT+0000 (UTC)
published: Mon Nov 20 2023 15:04:16 GMT+0000 (UTC)