Research map / Label-efficient and robust learning / Self-supervised and contrastive
Cross-Scale MAE: A Tale of Multiscale Exploitation in Remote Sensing
Maofeng Tang, Andrei Liviu Cozma, Konstantinos Georgiou, Hairong Qi
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 |
|---|---|---|---|
| add multi-scale augmentation (two-scale inputs) with reconstruction loss only (ablation) | MAE (SatMAE-like) · RESISC45 KNN 100% | KNN accuracy: 58.9 -> 69.2 | Improved |
| add cross-scale consistency (InfoNCE) loss in encoder (ablation) | Cross-Scale MAE with multi-scale input · RESISC45 KNN 100% | KNN accuracy: 69.2 -> 74.4 | Improved |
| add cross-scale prediction loss in decoder (ablation) | Cross-Scale MAE with multi-scale input · RESISC45 KNN 100% | KNN accuracy: 69.2 -> 76.5 | Improved |
| combine encoder consistency and decoder prediction losses (ablation) | Cross-Scale MAE · RESISC45 KNN 100% | KNN accuracy: 76.5 -> 79.3 | Improved |
| add negative samples in the encoder contrastive loss (none in decoder) (ablation) | Cross-Scale MAE · RESISC45 KNN 100% | KNN accuracy: 77.9 -> 79.3 | Improved |
| add negative samples in the decoder (ablation) | Cross-Scale MAE · RESISC45 KNN 100% | KNN accuracy: 79.3 -> 77.7 | Hurt |
| add negative samples in encoder in pure CL framework (ablation) | Cross-Scale MAE without reconstruction · RESISC45 KNN 100% | KNN accuracy: 58.7 -> 60.7 | Improved |
| add GSD positional encoding (ablation) | Cross-Scale MAE · RESISC45 KNN | KNN accuracy | Hurt |
| add GSD positional encoding or multi-scale augmentation to SatMAE (ablation) | SatMAE · RESISC45 KNN | KNN accuracy | Improved |
| replace Scale-MAE with Cross-Scale MAE | remote sensing MAE · UC Merced KNN (avg) | KNN accuracy: 75.0 -> 74.5 | Hurt |
| replace Scale-MAE with Cross-Scale MAE | remote sensing MAE · RESISC45 KNN (avg) | KNN accuracy: 70.0 -> 75.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Scale-MAE — Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning (ICCV 2023) · better on 8 of 10 shared results
- SatMAE — SatMAE: Pre-training Transformers for Temporal and Multi-Spectral Satellite Imagery (NeurIPS 2022) · better on 2 of 2 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
- Rethinking Transformers Pre-training for Multi-Spectral Satellite Imagery (CVPR 2024 · similarity 0.85)
- Galileo: Learning Global & Local Features of Many Remote Sensing Modalities (ICML 2025 · similarity 0.84)
- ChA-MAEViT: Unifying Channel-Aware Masked Autoencoders and Multi-Channel Vision Transformers for Improved Cross-Channel Learning (NeurIPS 2025 · similarity 0.84)
- SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery (CVPR 2024 · similarity 0.84)
- A Theoretical Analysis of Self-Supervised Learning for Vision Transformers (ICLR 2025 · similarity 0.84)
- Meta-Learning Hyperparameters for Parameter Efficient Fine-Tuning (CVPR 2025 · similarity 0.84)
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.
