Research map / Label-efficient and robust learning / Self-supervised and contrastive
InfMAE: A Foundation Model in The Infrared Modality
Fangcen Liu, Chenqiang Gao, Yaming Zhang, Junjie Guo, Jinghao Wang, Deyu Meng
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 information-aware masking and multi-scale encoder together (ablation) | InfMAE · MSRS-inf segmentation (UperNet) | mIoU (%): 71.4 -> 74.3 | Improved |
| add information-aware masking alone (ablation) | InfMAE · M3FD-inf detection (Mask R-CNN) | AP50 (%): 81.5 -> 78.9 | Hurt |
| add multi-scale encoder alone (ablation) | InfMAE · M3FD-inf detection (Mask R-CNN) | AP50 (%): 81.5 -> 86.3 | Improved |
| use shallow infrared decoder (depth 2) instead of deeper decoders (ablation) | InfMAE · MSRS-inf segmentation (UperNet) | mIoU (%): 72.9 -> 74.3 (vs depth 4) | Improved |
| increase masking sampling stride from 14 to 4 (ablation) | InfMAE · MSRS-inf segmentation (UperNet) | mIoU (%): 70.4 -> 74.3 | Improved |
| decrease masking stride from 4 to 2 (ablation) | InfMAE · MSRS-inf segmentation (UperNet) | mIoU (%): 74.3 -> 74.2 | No clear effect |
| extend pre-training from 400 to 1600 epochs (ablation) | InfMAE · MSRS-inf segmentation (UperNet) | mIoU (%): 74.3 -> 74.9 | Improved |
| extend pre-training from 400 to 1600 epochs (ablation) | InfMAE · M3FD-inf detection (Mask R-CNN) | mAP (%): 56.2 -> 56.0 | No clear effect |
| scale pre-training data from 100K (Inf10) to 300K (Inf30) (ablation) | InfMAE · MSRS-inf segmentation (UperNet) | mIoU (%): 69.1 -> 74.3 | Improved |
| scale pre-training data from curated 300K to unprocessed 500K (ablation) | InfMAE · MSRS-inf segmentation (UperNet) | mIoU (%): 74.3 -> 73.0 | Hurt |
| pre-train vanilla MAE on infrared Inf30 instead of visible IN1K-30 (ablation) | Vanilla MAE · MSRS-inf segmentation (UperNet) | mIoU (%): 66.7 -> 71.6 | Improved |
| self-supervised pre-training vs supervised training from scratch (ablation) | InfMAE encoder · MSRS-inf segmentation (UperNet) | mIoU (%): 61.1 -> 74.3 | Improved |
Later papers that compared with this method
1 later accepted paper reports InfMAE as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- DuGI-MAE: Improving Infrared Mask Autoencoders via Dual-Domain Guidance (AAAI 2026 · better on 8 of 9 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 Infrared Small Target Detection: A Foundation Driven Efficient Paradigm (NeurIPS 2026 · similarity 0.85)
- IF-Bench: Benchmarking and Enhancing MLLMs for Infrared Images with Generative Visual Prompting (CVPR 2026 · similarity 0.85)
- UNIP: Rethinking Pre-trained Attention Patterns for Infrared Semantic Segmentation (ICLR 2025 · similarity 0.84)
- FD2-Net: Frequency-Driven Feature Decomposition Network for Infrared-Visible Object Detection (AAAI 2025 · similarity 0.83)
- Semi-supervised Infrared Small Target Detection with Thermodynamic-Inspired Uneven Perturbation and Confidence Adaptation (AAAI 2025 · similarity 0.83)
- DifIISR: A Diffusion Model with Gradient Guidance for Infrared Image Super-Resolution (CVPR 2025 · similarity 0.82)
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.
