Conceptchannel attention
Channel Attention
An attention mechanism that learns a weight for each feature channel in order to emphasize important channels and suppress uninformative ones.
Timeline (last 90 days)
Oct 14
Channel attention (such as SE and ECA) learns which feature channels are important, outputs a channel-dimensional weight vector, and shares it across all spatial positions; spatial attention (such as the CBAM spatial branch and Non-local) learns which positions are important, outputs a spatial weight map, and shares it across all channels.
Unverified50%