Research map / Video / Video segmentation and tracking
Siamese Masked Autoencoders
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 asymmetric masking: leave the past frame unmasked and mask 95% of the future frame (ablation) | SiamMAE · DAVIS | J&F_m: 51.4 -> 58.1 | Helped |
| Increase asymmetric future-frame masking from 0.50 to 0.90 (ablation) | SiamMAE · DAVIS | J&F_m: 49.0 -> 58.4 | Helped |
| Use a siamese encoder instead of a joint encoder (ablation) | SiamMAE with cross-self decoder · DAVIS | J&F_m: 41.1 -> 58.1 | Helped |
| Use a cross-self decoder instead of a joint decoder (ablation) | Siamese encoder · DAVIS | J&F_m: 56.7 -> 58.1 | Helped |
| Use a cross-self decoder instead of a cross decoder (ablation) | Siamese encoder · DAVIS | J&F_m: 52.2 -> 58.1 | Helped |
| Add spatial augmentation (ablation) | SiamMAE · DAVIS | J&F_m: 56.8 -> 58.1 | Helped |
| Add color augmentation to spatial augmentation (ablation) | SiamMAE · DAVIS | J&F_m: 58.1 -> 56.7 | Hurt |
| Randomly sample frame gaps from 4-48 rather than use fixed gaps (ablation) | SiamMAE · DAVIS | J&F_m: 55.1 -> 58.1 | Helped |
| Use smaller ViT-S/8 patches rather than ViT-S/16 patches | SiamMAE · DAVIS-2017 | J&F_m: 62.0 -> 71.4 | Helped |
| Use SiamMAE rather than DINO | ViT-S/8 · VIP | mIoU: 39.5 -> 45.9 | Helped |
| Use SiamMAE rather than VFS | ViT-S/8 vs ResNet-50 · JHMDB | PCK@0.1: 60.9 -> 61.9 | Helped |
Closest papers by meaning
- Siamese Image Modeling for Self-Supervised Vision Representation Learning (CVPR 2023 · similarity 0.90)
- Fast Video Object Segmentation by Reference-Guided Mask Propagation (CVPR 2018 · similarity 0.88)
- VideoMAC: Video Masked Autoencoders Meet ConvNets (CVPR 2024 · similarity 0.87)
- Adaptive Siamese Masked Autoencoder with Global Optimization for Unsupervised Point Cloud Shape Correspondence (AAAI 2025 · similarity 0.87)
- Self-Supervised Cross-View Correspondence with Predictive Cycle Consistency (CVPR 2025 · similarity 0.86)
- Closest Neighbors are Harmful for Lightweight Masked Auto-encoders (CVPR 2025 · similarity 0.86)
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.