CLIP(CNN)

[chenpan0103]

Excuse me, the existing code only gives the code about ALBEF, TCL, CLIP(VIT), I did not find CLIP(CNN), what should I do if I want to use it?

[Sammy42779]

Hello, if you want to use CLIP(CNN), directly change the "image_encoder" from "ViT-B/16" to "RN101" since we use ResNet-101 as the base architecture for the image encoder (Sec. 5.1).

[chenpan0103]

Thanks for your reply, it does work. By the way, I have one more question. How can I reproduce the results of attack methods other than SGA in the table? Only the SGA code seems to have been given

[chenpan0103]

Sorry to bother you, it seems that the code from CLIP to TCL or ALBEF is missing. I would appreciate it if you could provide it.

[Zoky-2020]

Thanks for your reply, it does work. By the way, I have one more question. How can I reproduce the results of attack methods other than SGA in the table? Only the SGA code seems to have been given

We will update the related codes soon. Here is the official repo. of Co-Attack, https://github.com/adversarial-for-goodness/Co-Attack. PGD and BERT-Attack are modified based on 'Towards Deep Learning Models Resistant to Adversarial Attacks (https://arxiv.org/abs/1706.06083)' and 'BERT-ATTACK: Adversarial Attack Against BERT Using BERT' (https://arxiv.org/abs/2004.09984), respectively. And Sep-Attack denotes the combination of PGD and BERT-Attack. All the official repositories are accessible.

[Zoky-2020]

Sorry to bother you, it seems that the code from CLIP to TCL or ALBEF is missing. I would appreciate it if you could provide it.

Thanks for your interest in our work. The code towards clip2albef is available now; see in eval_clip-vit2albef_flickr.py

[chenpan0103]

Sorry to bother you, it seems that the code from CLIP to TCL or ALBEF ，is missing. I would appreciate it if you could provide it.

Thanks for your interest in our work. The code towards clip2albef is available now; see in eval_clip-vit2albef_flickr.py

Thank you very much for your efforts to update the code, looking forward to your subsequent code updates. In addition, I would appreciate it if you could upload TCL2ALBEF ，ALBEF2TCL， CLIP2CLIP in SGA first when you are free. Thanks again for your reply.