A Comparative Study of Transformer and Convolutional Models for Crop Segmentation from Satellite Image Time Series
Abstract
Crop segmentation from satellite image time series (SITS) is a fundamental task for agricultural monitoring and land-use analysis. While convolutional neural networks (CNNs) have been widely used, transformer-based architectures offer alternative mechanisms for representing spatial and temporal dependencies in multispectral data. This paper presents a comparative study of CNN and transformer-based segmentation models for crop mapping from Sentinel-2 time series, including 3D U-Net, 3D FPN, 3D DeepLabv3, and three transformer architectures: Swin UNETR, TSViT, and VistaFormer, which adopt different strategies for capturing temporal dependencies. Experiments on the Munich and Lombardia datasets show that TSViT achieves the best overall results, slightly surpassing 3D U-Net, which remains a strong CNN baseline. VistaFormer offers the best efficiency, while Swin UNETR performs competitively but is less effective than transformers that explicitly model temporal dynamics. These results highlight that temporal modelling is critical for SITS: TSViT outperforms CNNs and approaches that treat time as an additional spatial dimension, while VistaFormer provides a strong efficiency-performance trade-off.
Keywords
Cite
@article{arxiv.2412.01944,
title = {A Comparative Study of Transformer and Convolutional Models for Crop Segmentation from Satellite Image Time Series},
author = {Mattia Gatti and Ignazio Gallo and Nicola Landro and Christian Loschiavo and Anwar Ur Rehman and Mirco Boschetti and Riccardo La Grassa},
journal= {arXiv preprint arXiv:2412.01944},
year = {2026}
}
Comments
This version corrects an error in the evaluation pipeline affecting previously reported metrics. Results have been recomputed, leading to updated values and a revised conclusion: the adapted Swin UNETR model does not outperform CNN baselines. Tables, figures, and comparisons have been updated, and the analysis has been extended to include additional transformer-based models