**ScheduledDropPath** is a modified version of [DropPath](https://paperswithcode.com/method/droppath). In DropPath, each path in the cell is stochastically dropped with some fixed probability during training. In ScheduledDropPath, each path in the cell is dropped out with a probability that is linearly increased over the course of training.

**DFDNet**, or **DFDNet**, is a deep face dictionary network for face restoration to guide the restoration process of degraded observations. Given a LQ image $I\_{d}$, the DFDNet selects the dictionary features that have the most similar structure with the input. Specially, we re-norm the whole dictionaries via component AdaIN (termed as CAdaIN) based on the input component to eliminate the distribution or style diversity. The selected dictionary features are then utilized to guide the restoration process via dictionary feature transformation.

DFDNet

Blind Face Restoration via Deep Multi-scale Component Dictionaries

ScheduledDropPath

Learning Transferable Architectures for Scalable Image Recognition

The **Routing Transformer** is a [Transformer](https://paperswithcode.com/method/transformer) that endows self-attention with a sparse routing module based on online k-means. Each attention module considers a clustering of the space: the current timestep only attends to context belonging to the same cluster. In other word, the current time-step query is routed to a limited number of context through its cluster assignment.

Source	Learning Transferable Architectures for Scalable Image Recognition
Year	2000
Data Source	CC BY-SA - https://paperswithcode.com