How do I implement Figure2(b) in detail?

xxxnell / how-do-vits-work

(ICLR 2022 Spotlight) Official PyTorch implementation of "How Do Vision Transformers Work?"

https://arxiv.org/abs/2202.06709

Apache License 2.0

806 stars 79 forks source link

How do I implement Figure2(b) in detail? #43

Closed pi-wo closed 4 months ago

pi-wo commented 4 months ago

I have obtained the performance evaluation results of resnet and vit under different types of noise and different intensification using the robustness code you have given. But I do not know how the results of the effects of different frequencies of noise on the model performance in Figure2(b) are obtained. Can you share the specific code?

xxxnell commented 4 months ago

Hi @pi-wo,

Thank you for reaching out. Since Figure 2b depends on (ImageNet) datasets, I haven't uploaded the code. Please leave your email address or send a message to namuk.park@gmail.com, and I'll send you the snippet for the experiment.

Hope you have a nice weekend!

pi-wo commented 4 months ago

Thank you for your reply to paper Figure2(b), which is of great help to my research. My email is dj2650898330@gmail.com.

Have a nice weekend.