Research map / Vision-language and multimodal / Cross-modal learning
MultiMAE: Multi-modal Multi-task Masked Autoencoders
Roman Bachmann, David Mizrahi, Andrei Atanov, Amir Zamir
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 ground-truth depth as extra input (RGB-D) at transfer | MultiMAE · Hypersim semantic segmentation | mIoU: 37.0 -> 47.6 | Improved |
| add ground-truth depth as extra input (RGB-D) at transfer | MAE · Hypersim semantic segmentation | mIoU: 36.5 -> 36.9 | No clear effect |
| add ground-truth depth as extra input (RGB-D) at transfer | MultiMAE · NYUv2 semantic segmentation | mIoU: 52.0 -> 56.0 | Improved |
| add ground-truth depth as extra input (RGB-D) at transfer | MAE · NYUv2 semantic segmentation | mIoU: 50.8 -> 49.3 | Hurt |
| add pseudo-labeled depth and semantic segmentation inputs | MultiMAE · NYUv2 semantic segmentation | mIoU: 52.0 -> 54.0 | Improved |
| add pseudo-labeled depth and semantic segmentation inputs | MultiMAE · Hypersim semantic segmentation | mIoU: 37.0 -> 40.1 | Improved |
| add pseudo-labeled depth and semantic segmentation inputs | MAE · NYUv2 semantic segmentation | mIoU: 50.1 -> 49.3 | Hurt |
| add pseudo-labeled depth and semantic segmentation inputs | MultiMAE · ADE20K semantic segmentation | mIoU: 46.2 -> 47.1 | Improved |
| add semantic segmentation as pre-training task alongside RGB-depth (ablation) | MultiMAE (400 epochs) · NYUv2 semantic segmentation | mIoU: 45.8 -> 50.6 | Improved |
| add depth as pre-training task alongside RGB-semseg (ablation) | MultiMAE (400 epochs) · NYUv2 semantic segmentation | mIoU: 51.6 -> 50.6 | Hurt |
| add masking to multi-task RGB-depth-semseg pre-training (ablation) | MultiMAE (400 epochs) · NYUv2 semantic segmentation | mIoU: 48.6 -> 50.6 | Improved |
| use deeper decoder (depth 8 instead of depth 2) in MAE (ablation) | MAE · ADE20K | mIoU: 43.3 -> 46.2 | Improved |
Later papers that compared with this method
1 later accepted paper reports MultiMAE as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Multi-Modal Contrastive Masked Autoencoders: A Two-Stage Progressive Pre-training Approach for RGBD Datasets (CVPR 2025 · 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
- ColorMAE: Exploring data-independent masking strategies in Masked AutoEncoders (ECCV 2024 · similarity 0.89)
- Multi-Modal Contrastive Masked Autoencoders: A Two-Stage Progressive Pre-training Approach for RGBD Datasets (CVPR 2025 · similarity 0.89)
- Understanding and Enhancing Mask-Based Pretraining towards Universal Representations (NeurIPS 2025 · similarity 0.88)
- CrossMAE: Cross-Modality Masked Autoencoders for Region-Aware Audio-Visual Pre-Training (CVPR 2024 · similarity 0.87)
- R-MAE: Regions Meet Masked Autoencoders (ICLR 2024 · similarity 0.87)
- The Dynamic Duo of Collaborative Masking and Target for Advanced Masked Autoencoder Learning (AAAI 2025 · similarity 0.87)
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.
