Research map / Vision-language and multimodal / Human and embodied modeling
SemMAE: Semantic-Guided Masking for Learning Masked Autoencoders
Gang Li, Heliang Zheng, Daqing Liu, Chaoyue Wang, Bing Su, Changwen Zheng
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 semantic-guided masking (learned parts) to MAE | MAE ViT-Base 16x16 · ImageNet-1k | linear probing accuracy: 63.7 -> 65.0 | Improved |
| add semantic-guided masking to MAE with 8x8 patches | MAE ViT-Base 8x8 · ImageNet-1k | linear probing accuracy: 66.8 -> 68.7 | Improved |
| use 8x8 instead of 16x16 patch size (ablation) | SemMAE · ImageNet-1k | fine-tuning accuracy: 83.34 -> 84.50 | Improved |
| use 8x8 instead of 16x16 patch size (ablation) | MAE · ImageNet-1k | fine-tuning accuracy: 83.26 -> 83.10 | Hurt |
| use semantic parts from iBOT attention instead of baseline random masking (ablation) | MAE · ImageNet-1k | linear probing accuracy: 63.7 -> 63.6 | No clear effect |
| use self-supervised refined part learning instead of iBOT-initialized parts (ablation) | SemMAE · ImageNet-1k | linear probing accuracy: 63.6 -> 65.0 | Improved |
| mask 75% of whole parts instead of random masking (ablation) | MAE · ImageNet-1k | linear probing accuracy: 66.8 -> 52.9 | Hurt |
| mask 75% patches in each part instead of random masking (ablation) | MAE · ImageNet-1k | linear probing accuracy: 66.8 -> 66.5 | No clear effect |
| easy-to-hard adaptive masking (patches per part to whole parts) instead of random masking (ablation) | MAE · ImageNet-1k | linear probing accuracy: 66.8 -> 68.7 | Improved |
| reverse schedule (hard to easy, parts first) instead of easy to hard (ablation) | SemMAE · ImageNet-1k | linear probing accuracy: 68.7 -> 63.3 | Hurt |
| gamma 2 instead of 1/3 in masking schedule (ablation) | SemMAE · ImageNet-1k | linear probing accuracy: 66.2 -> 68.7 | Improved |
| gamma 3 instead of gamma 2 (ablation) | SemMAE · ImageNet-1k | linear probing accuracy: 68.7 -> 68.6 | No clear effect |
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.
- BeiT — BEiT: BERT Pre-Training of Image Transformers (ICLR 2022) · better on 1 of 1 shared result
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)
- R-MAE: Regions Meet Masked Autoencoders (ICLR 2024 · similarity 0.89)
- Understanding and Enhancing Mask-Based Pretraining towards Universal Representations (NeurIPS 2025 · similarity 0.87)
- The Dynamic Duo of Collaborative Masking and Target for Advanced Masked Autoencoder Learning (AAAI 2025 · similarity 0.87)
- Self-Guided Masked Autoencoder (NeurIPS 2024 · similarity 0.87)
- Closest Neighbors are Harmful for Lightweight Masked Auto-encoders (CVPR 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.
