tencentarc / desra Goto Github PK

According to my understanding, you should fine-tune RealESRGAN and other models on the synthetic artifact training set to eliminate artifacts. Can you provide the final checkpoints?

Hello,there is a issue have generated in your coding''from mmseg.apis import inference_segmentor, init_segmentor, show_seg
from mmseg.core.evaluation import get_palette'', because the mmseg do not have this files. I have check all method in mmseg files,these package do not exist.
secondly, How to use your script/artifact_dection to clear the Artifact in my own image generated by another GAN?
Thank you very much!(The demo to clear one image)

Hello, I would like to apply a new image to DeSRA. I know how to create GAN-SR and MSE-SR images, but I don't know how to create GT-Mask and DeSRA-Mask images. I would be very grateful if you could explain. :)

Hi, I have encoutered a problem while loading the checkpoint of Segformer model. Specifically, the model and loaded state dict do not match exactly. The error messages are as below:

Loads checkpoint by local backend from path: checkpoints/mit_b5_20220624-658746d9.pth
The model and loaded state dict do not match exactly
unexpected key in source state_dict:.....
missing keys in source state_dict:...
Traceback (most recent call last):
File "demo/artifact_detection.py", line 320, in
main()
File "demo/artifact_detection.py", line 215, in main
model = init_model(args.config, args.checkpoint, device=args.device)
File "mmsegmentation/mmseg/apis/inference.py", line 62, in init_model
dataset_meta = checkpoint['meta'].get('dataset_meta', None)
KeyError: 'meta'

Would you please provide the exact Segformer checkpoint that works? Thanks a lot!

Hello, I encountered the following problem, from mmseg.apis import inference_segmentor, init_segmentor, show_seg
ImportError: cannot import name 'inference_segmentor' from 'mmseg.apis' . How to solve it.

Installing the package according to the version you specified will not run the code correctly, causing some errors. Please fix the code or version dependencies.

In artifact_detection.py
I think
Line 10 "from mmseg.apis import inference_segmentor, init_segmentor, show_seg from mmseg.apis import inference_model, init_model, show_result_pyplot" should be change to
"from mmseg.apis import inference_model, init_model, show_result_pyplot from mmseg.core.evaluation import get_palette"
because of function name in mmsegmentation.

What is your opinion on my thoughts?

Which MSE model do you use? I didn't find this model in the paper. Is it EDSR or SRResnet?
Thank you for your reply！

请问checkpoints文件在哪下载？

I created Final_Artifact_Map with Urban100 and BSD100 datasets, but there are only few files that have been segmented. Why is that so? I wonder if it is due to because of the difference in resolution between the test dataset (DeSRA) and the dataset I used(Urban100, BSD100)
or because segmentation is not suitable. (config and checkpoints set default)

Hello, I tried to reproduce your results, but after correcting many problems, it still only completed the mask segmentation and showed that the array was out of bounds when loading the category. Is this problem caused by an error in the class name in the class?
“ File "demo/artifact_detection.py", line 131, in main
label_name = CLASSES[label]
IndexError: list index out of range”
Can you create a colab demo in your project？Thank you！