Research map / Label-efficient and robust learning / Self-supervised and contrastive
MOMO: Mars Orbital MOdel Foundation Model for Mars Orbital Applications
Mirali Purohit, Bimal Gajera, Irish Mehta, Bhanu Tokas, Jacob Adler, Steven Lu, Scott Dickenshied, Serina Diniega, Brian Bue, Umaa Rebbapragada, Hannah Kerner
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use EVL checkpoint selection instead of early-stopping checkpoints (ablation) | MOMO merge · Mars-Bench Boulder | mIoU: 0.12 -> 0.20 | Improved |
| use EVL checkpoint selection instead of last-epoch checkpoints (ablation) | MOMO merge · Mars-Bench Boulder | mIoU: 0.18 -> 0.20 | Improved |
| use EVL checkpoint selection instead of ES/LE (ablation) | MOMO merge · Mars-Bench ConeQuest | mIoU: 0.68/0.70 -> 0.71 | No clear effect |
| use EVL checkpoint selection instead of ES/LE (ablation) | MOMO merge · Mars-Bench Crater Binary | mIoU: 0.50 -> 0.54 | Improved |
| merge per-sensor models instead of joint data pretraining (ablation) | MOMO · Mars-Bench Frost | F1: 0.44 -> 0.98 | Improved |
| merge per-sensor models instead of joint data pretraining (ablation) | MOMO · Mars-Bench segmentation tasks | mIoU: MOMO outperforms DM by 5.4% in segmentation | Improved |
| use merged multi-sensor model instead of single-sensor pretraining (ablation) | MOMO · Mars-Bench segmentation | mIoU: +4.2% over same-sensor | Improved |
| same-sensor vs cross-sensor pretraining (ablation) | sensor-specific models · Mars-Bench | F1/mIoU | Improved |
| use MOMO instead of ImageNet pretraining | MOMO · Mars-Bench classification | F1: ~1.25% average F1 gain | Improved |
| use MOMO instead of ImageNet pretraining | MOMO · Mars-Bench Landmark | F1: 0.92 -> 0.91 | Hurt |
| use MOMO instead of ImageNet pretraining | MOMO · Mars-Bench segmentation | mIoU: ~1% average mIoU gain; Crater Binary 0.55 -> 0.54 | Mixed |
| use MOMO instead of random init | MOMO · Mars-Bench DoMars16k | F1: 0.73 -> 0.92 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SatMAE — SatMAE: Pre-training Transformers for Temporal and Multi-Spectral Satellite Imagery (NeurIPS 2022) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Mars-Bench: A Benchmark for Evaluating Foundation Models for Mars Science Tasks (NeurIPS 2025 · similarity 0.86)
- M3LEO: A Multi-Modal, Multi-Label Earth Observation Dataset Integrating Interferometric SAR and Multispectral Data (NeurIPS 2024 · similarity 0.81)
- Martian World Model: Controllable Video Synthesis with Physically Accurate 3D Reconstructions (NeurIPS 2025 · similarity 0.80)
- REOBench: Benchmarking Robustness of Earth Observation Foundation Models (NeurIPS 2025 · similarity 0.80)
- RoMA: Scaling up Mamba-based Foundation Models for Remote Sensing (NeurIPS 2025 · similarity 0.80)
- Towards Realistic Earth-Observation Constellation Scheduling: Benchmark and Methodology (NeurIPS 2025 · similarity 0.79)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
