Jianyi Wang

icon indicating a website link https://iceclear.github.io icon indicating an email [email protected]

icon company Nanyang Technological University (NTU) icon location Singapore icon location Ph.D. candidate from MMLab@NTU, mainly focusing on low-level computer vision.

Results 157 comments of Jianyi Wang

Stop at validation sanity check

> Hi. Please refer to #44 . here

Stop at validation sanity check

Thanks for the suggestion. I may improve this when I am avaliable.

AttributeError: module 'ldm.models.diffusion.ddpm' has no attribute 'LatentDiffusionSRTextWT'

It should be an env problem. Try `pip install -e .`

RuntimeError: PytorchStreamReader in demo colab StableSR_inference.ipynb

Maybe this is similar to your question? https://stackoverflow.com/questions/71617570/pytorchstreamreader-failed-reading-zip-archive-failed-finding-central-directory

what is the command to run the upscaler on colab

I think you just need to run the code line by line.

Fine-tuning stable diffusion

Hi～our approach does not need to finetune the SD part. We directly use the official checkpoint. You can find the link in readme.

Fine-tuning stable diffusion

> Thanks for the reply! I see in your paper in experiment detail part that mention fine tune the diffusion model of stableSR, so that means fine tune the SFT...

4x upscale is fine but 2x upsacle poor with my own finetuned model

I am not sure about what happens but you can just downsample the 4x results to get the 2x one.

During the model training process, the degraded images in the training set were restored to the ink style.

Hi. The main problem should be related to your own training dataset. I think you need to adjust the degradation accordingly or change the training dataset.

How long does it take to generate training data？

The current code does not support multiple GPUs. But you can run several processes one by one at the same time. I used to run the code using about 12...