AttributeError: 'EasyDict' object has no attribute 'model'

[JD-2006]

Here is my argument: python inference_image.py configs/base_dvd.yaml demo/dbn_ckpt.pth --img=2016-3A-00000.png --save_path=inference_results

Traceback (most recent call last): File "C:\Users\Test\Documents\Image_Scripts\SimDeblur\inference_image.py", line 81, in inference() File "C:\Users\Test\Documents\Image_Scripts\SimDeblur\inference_image.py", line 46, in inference arch = build_meta_arch(cfg) File "C:\Users\Test\Documents\Image_Scripts\SimDeblur\simdeblur\model\build.py", line 38, in build_meta_arch ret = META_ARCH_REGISTRY.get(name)(cfg) File "C:\Users\Test\Documents\Image_Scripts\SimDeblur\simdeblur\model\meta_arch\plain_cnn.py", line 40, in init self.model = self.build_model(cfg).to(self.device) File "C:\Users\Test\Documents\Image_Scripts\SimDeblur\simdeblur\model\meta_arch\base_arch.py", line 76, in build_model model = build_backbone(cfg.model) AttributeError: 'EasyDict' object has no attribute 'model'

[ljzycmd]

Hi @JD-2006, with the above details, you may use the wrong config file during inference. Note that configs/base_dvd.yaml is a config for DVD video deblurring dataset, not including the model config.

You should use a config file that contains a specified model configuration for single image deblurring, e.g., config yaml in folder configs/srn, configs/minmounet.

[JD-2006]

Thanks that fixed it!

p.s. Any chance of a Restormer yaml?

[ljzycmd]

Hi @JD-2006, I have updated the configs for the Restormer model. Please check.