KAN You See It? KANs and Sentinel for Effective and Explainable Crop Field Segmentation

TL;DR

This paper introduces KAN layers integrated into U-Net for crop field segmentation, achieving improved accuracy and explainability using Sentinel satellite data, with a focus on boundary detection and channel relevance.

Contribution

It presents the novel integration of Kolmogorov-Arnold networks (KANs) into U-Net for enhanced crop segmentation and explainability with satellite imagery.

Findings

2% IoU improvement over traditional U-Net

High plausibility of gradient-based explanations

Identification of irrelevant channels

Abstract

Segmentation of crop fields is essential for enhancing agricultural productivity, monitoring crop health, and promoting sustainable practices. Deep learning models adopted for this task must ensure accurate and reliable predictions to avoid economic losses and environmental impact. The newly proposed Kolmogorov-Arnold networks (KANs) offer promising advancements in the performance of neural networks. This paper analyzes the integration of KAN layers into the U-Net architecture (U-KAN) to segment crop fields using Sentinel-2 and Sentinel-1 satellite images and provides an analysis of the performance and explainability of these networks. Our findings indicate a 2\% improvement in IoU compared to the traditional full-convolutional U-Net model in fewer GFLOPs. Furthermore, gradient-based explanation techniques show that U-KAN predictions are highly plausible and that the network has a very high ability to focus on the boundaries of cultivated areas rather than on the areas themselves. The per-channel relevance analysis also reveals that some channels are irrelevant to this task.