FAN
FAN copied to clipboard
NVlabs
Official PyTorch implementation of Fully Attentional Networks
I found you have implemented a SAM-based ApexScaler in [here](https://github.com/NVlabs/FAN/blob/6467085dbc6a410a616e17a36fdfa375d21770bc/utils/scaler.py#L10) Is it useable now? Can we use this SAM-based ApexScaler to train FAN?
Hi, Thanks for your excellent work. What are the differences in results on ImageNet-C between Table 11 and Table 8? For example, with FAN-L-Hybrid, It is 68.3 in Table 8...
Hi there I receive the following error when running the forward method on models with `FANBlock_SE` `TypeError: forward() missing 2 required positional arguments: 'H' and 'W'` An example of code...
When i running detection code use dist_train.sh on my machine, TypeError: CascadeRCNN: fan_small_12_p4_hybrid: __init__() got an unexpected keyword argument 'default_cfg' was occur. My conda env use mmcv==1.3.0 and mmdet==2.11.0. I...
Hi, Thank you for sharing the code and model weights for your work. Would it be possible to clarify if the shared weights correspond to the EMA model or not...
Hi, Could you please provide more detailed instructions about dataset? I tried to download the clean Imagenet dataset, but there are many download links. I also tried to download Imagenet-c,...
https://github.com/NVlabs/FAN/blob/ee1b7df1016205ed26e28c0683b14238f3530a84/models/fan.py#L311-L315 Hi! Thank you for your great work! According to CaiT, I think the code should be in the following form: `cls_token = x[:, 0:1] + self.drop_path(self.gamma2 * self.mlp(x[:, 0:1]...
Thank you for you great job. When I read the paper, I am confused about the "Attentional feature transformation" part. I don't know why the channel dimension attention can stand...
So, the target is to train on the clean dataset, like ImageNet, and show robustness on the corrupt dataset, like ImageNet-C, right?
# Welcome update to OpenMMLab 2.0 I am Vansin, the technical operator of OpenMMLab. In September of last year, we announced the release of OpenMMLab 2.0 at the World Artificial...
