How to visualize the attention map of the cross-attention

Atten4Vis / ConditionalDETR

This repository is an official implementation of the ICCV 2021 paper "Conditional DETR for Fast Training Convergence". (https://arxiv.org/abs/2108.06152)

Apache License 2.0

358 stars 48 forks source link

How to visualize the attention map of the cross-attention #4

Closed GWwangshuo closed 2 years ago

GWwangshuo commented 3 years ago

Thanks for your nice work. Could you also please provide the code of visualizing the attention map of the cross-attention? Thanks a lot.

DeppMeng commented 2 years ago

Duplicated with issue #5.