NAFNet

NAFNet copied to clipboard

I wanted to use this model on my application and wanted a tflite transformed model. How can I convert this model into tflite any iputs will be highly appreciated. Thank...

adkbbx

Hi! I've converted [NAFNet-REDS-width64](https://github.com/megvii-research/NAFNet#results-and-pre-trained-models) to CoreML. If you're interested in CoreML weights or inference code, you can checkout this repository: https://github.com/Vadbeg/nafnet-coreml

Vadbeg

I am trying to run the NAFNet pre-trained model for deblurring images, I was able to run about 5 images before I got the error "Killed: 9", Now that I...

jacobstasiewicz

您好，我对您的工作很感兴趣，想复现一下，但是我只有两张卡，如果要训练的话， batch_size_per_gpu，lr和total_iter这几个参数要怎么修改才能达到论文的效果呢

77loggedaloha

Hi, thanks for your work. I found that the biases of your decoder block are set to false. is there any reason for that ? thanks!

myoh97

请问能提供在REDS数据集上训练的通道数为32的预训练模型吗

77loggedaloha

使用两颗GPU复现NAFNet-width32疑问

1

作者您好 我在先前的提问中看到作者是用8颗GPU训练，而当使用2颗GPU训练时，是否要将training iteraation增加到原始设定的4倍呢? 我尝试将training iteration增加4倍，但达不到paper中GoPro dataset的数字 (差0.3dB) 不过把iteration数增加到8倍后，就超过paper当中的数字 (多约0.2 dB) 不太确定怎么样的设定才是对的

Barry-CHIU

strange output in denoising model

1

When the input image is of any size and the color value is rgb (33, 25, 23), strange output will be obtained.  

fxjy15550

Train with 3 GPUs

3

您好， 我想问一下要是分三张卡跑该怎么修改代码呢 简单地将`python -m torch.distributed.launch --nproc_per_node=8 --master_port=4321 basicsr/train.py -opt options/train/NAFSSR/NAFSSR-B_x4.yml --launcher pytorch`里的--nproc_per_node改成3貌似不行 想问一下您具体的enable multiple gpus on torch的代码在哪里呢 谢谢

yunzhelimac

Hello How are you? Thanks for contributing to this project. I am going to train a model on my custom dataset. There are different sizes of validation images. I met...

rose-jinyang