Computer Vision
掩码图像建模
Masked Image ModelingA self-supervised learning framework where input data (e.g., images) is randomly masked, and the model is trained to reconstruct the original data. / 一种自监督学习框架，通过随机掩码输入数据（如图像）并训练模型重建原始数据。
- Knowledge type
- 数学理论
- First proposed or discovered
- Proposer
- Turkey Farm member
- Creator
- 农场主
- Last updated
- 2026-11-11 19:06:37
Sub-knowledge points
Patch maskingFeature maskingReconstruction head design
Higher-order knowledge points
Self-supervised learning / 数学理论Contrastive learning
Main parameters and performance indicators
masking_ratio: Proportion of pixels or patches randomly selected for maskingreconstruction_loss: Metric used to measure the difference between the original and reconstructed data
Open problems
No unresolved problem recorded yet
Related literature
ranked by community starsWhen Grouped Cyclic Shift meets masked image modeling: Effective pre-training for data-scarce 3D ultrasound analysis tasksOpen source
Related products and links
No item recorded yet
Creation and modification trail
1 recordsRecord title: Masked Image Modeling.
Node update feedback
review required