Open Bugs-Bunny01 opened 1 year ago
About the attention visualization, you can take a look at those repositories. https://github.com/yiyixuxu/TimeSformer-rolled-attention/blob/main/visualize_attn_util.py https://github.com/jacobgil/vit-explain/tree/main We mainly change our code Based on them.
We select to use ResNet18 as the pre-trained model for the baseline basic_CNN.
Hi, thank you for contributing to the robotics open source project. I am reproducing your paper, but I do not understand the attention visualization. How do you do it? And how about imagenet pre-trained models? Can I ask you to open source the code, I am a beginner, do not understand. Thank you again.