Published May 15, 2025 | Version v1
Dataset Open

Mars-Bench: mb-boulder_det

  • 1. Arizona State University
  • 2. Jet Propulsion Laboratory, California Institute of Technology

Description

This is the object detection version of the Boulder dataset, designed to localize boulders on the Martian surface using high-resolution orbital imagery from the HiRISE camera. Each image is annotated with manually curated bounding boxes that delineate individual boulders, with annotations created by planetary scientists to ensure high-quality and scientifically accurate labels. The dataset includes predefined train, validation, and test splits. This is one of the smallest datasets in Mars-Bench with only tens of samples. Given the small-object nature of the task, this dataset presents a valuable benchmark for evaluating object detection models in low-data regimes, a setting of growing interest in the computer vision community.

 

To explore all Mars-Bench resources, visit our project website:

https://mars-bench.github.io/

Files

data.zip

Files (9.7 MB)

Name Size Download all
md5:8f4ea80c6ec213c8b019bb4bbbecb7f5
9.7 MB Preview Download
md5:fb4b09a67195f0876bdc64a4c4231222
160 Bytes Download
md5:fdf56e0579c880d632048c1b4e2a007f
1.4 kB Preview Download

Additional details