# SenCLIP: Enhancing zero-shot land-use mapping for Sentinel-2 with ground-level prompting

Source: [SenCLIP: Enhancing zero-shot land-use mapping for Sentinel-2 with ground-level prompting](<https://www.sota2.com/research/paper/68fa7aae1e209131cea8495d>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T05:25:52.609Z

Authors: Pallavi Jain, Dino Ienco, Roberto Interdonato, Tristan Berchoux, Diego Marcos

Published: 2024-12-11T16:52:14.000Z

## Abstract

Pre-trained vision-language models (VLMs), such as CLIP, demonstrate impressive zero-shot classification capabilities with free-form prompts and even show some generalization in specialized domains. However, their performance on satellite imagery is limited due to the underrepresentation of such data in their training sets, which predominantly consist of ground-level images. Existing prompting techniques for satellite imagery are often restricted to generic phrases like a satellite image of ..., limiting their effectiveness for zero-shot land-use and land-cover (LULC) mapping. To address these challenges, we introduce SenCLIP, which transfers CLIPs representation to Sentinel-2 imagery by leveraging a large dataset of Sentinel-2 images paired with geotagged ground-level photos from across Europe. We evaluate SenCLIP alongside other SOTA remote sensing VLMs on zero-shot LULC mapping tasks using the EuroSAT and BigEarthNet datasets with both aerial and ground-level prompting styles. Our approach, which aligns ground-level representations with satellite imagery, demonstrates significant improvements in classification accuracy across both prompt styles, opening new possibilities for applying free-form textual descriptions in zero-shot LULC mapping.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Ground-to-Satellite Retrieval on Overall](<https://www.sota2.com/research/sota/ground-to-satellite-retrieval-on-overall.md>): Ground-to-Satellite Retrieval; Overall; Recall@1: 36.2.

- [Ground-to-Satellite Retrieval on Land Cover](<https://www.sota2.com/research/sota/ground-to-satellite-retrieval-on-land-cover.md>): Ground-to-Satellite Retrieval; Land Cover; Recall@1: 48.3.

- [Ground-to-Satellite Retrieval on Land Use](<https://www.sota2.com/research/sota/ground-to-satellite-retrieval-on-land-use.md>): Ground-to-Satellite Retrieval; Land Use; Recall@1: 58.9.

- [Satellite-to-Ground Retrieval on Overall](<https://www.sota2.com/research/sota/satellite-to-ground-retrieval-on-overall.md>): Satellite-to-Ground Retrieval; Overall; Recall@1: 30.6.

- [Scenicness Estimation on Scenicness Dataset](<https://www.sota2.com/research/sota/scenicness-estimation-on-scenicness-dataset.md>): Scenicness Estimation; Scenicness Dataset; Pearson's R: 0.421.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2412.08536>)

[PDF](<https://arxiv.org/pdf/2412.08536.pdf>)
