# CraterBench-R: Instance-Level Crater Retrieval for Planetary Scale

Source: [CraterBench-R: Instance-Level Crater Retrieval for Planetary Scale](<https://www.sota2.com/research/paper/69d732667d0172140ec76989>)

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

Data updated: 2026-04-10T05:00:04.898Z

Authors: Jichao Fang, Lei Zhang, Michael Phillips, Wei Luo

Published: 2026-04-06T03:04:31.000Z

## Abstract

Impact craters are a cornerstone of planetary surface analysis. However, while most deep learning pipelines treat craters solely as a detection problem, critical scientific workflows such as catalog deduplication, cross-observation matching, and morphological analog discovery are inherently retrieval tasks. To address this, we formulate crater analysis as an instance-level image retrieval problem and introduce CraterBench-R, a curated benchmark featuring about 25,000 crater identities with multi-scale gallery views and manually verified queries spanning diverse scales and contexts. Our baseline evaluations across various architectures reveal that self-supervised Vision Transformers (ViTs), particularly those with in-domain pretraining, dominate the task, outperforming generic models with significantly more parameters. Furthermore, we demonstrate that retaining multiple ViT patch tokens for late-interaction matching dramatically improves accuracy over standard single-vector pooling. However, storing all tokens per image is operationally inefficient at a planetary scale. To close this efficiency gap, we propose instance-token aggregation, a scalable, training-free method that selects K seed tokens, assigns the remaining tokens to these seeds via cosine similarity, and aggregates each cluster into a single representative token. This approach yields substantial gains: at K=16, aggregation improves mAP by 17.9 points over raw token selection, and at K=64, it matches the accuracy of using all 196 tokens with significantly less storage. Finally, we demonstrate that a practical two-stage pipeline, with single-vector shortlisting followed by instance-token reranking, recovers 89-94% of the full late-interaction accuracy while searching only a small candidate set. The benchmark is publicly available at hf.co/datasets/jfang/CraterBench-R.

## 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.

- [Image Retrieval on Curated-5K (test)](<https://www.sota2.com/research/sota/image-retrieval-on-curated-5k-test.md>): Image Retrieval; Curated-5K (test); Recall@1: 37.4.

- [Instance Retrieval on CraterBench-R](<https://www.sota2.com/research/sota/instance-retrieval-on-craterbench-r.md>): Instance Retrieval; CraterBench-R; R@1: 37.4.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2604.06245>)

[PDF](<https://arxiv.org/pdf/2604.06245.pdf>)