image-super-resolution
image-super-resolution copied to clipboard
idealo
🔎 Super-scale your images and run experiments with Residual Dense and Adversarial Networks.
In the training object, the default losses are MSE, binary cross-entropy, and MAE. Is this what the models are minimizing? If so, how can we make the losses the same...
`ERROR: Cannot install isr==1.9.0, isr==1.9.1, isr==1.9.2, isr==1.9.3, isr==2.0.0, isr==2.0.1, isr==2.0.2, isr==2.0.5, isr==2.1, isr==2.1.1 and isr==2.2.0 because these package versions have conflicting dependencies.`
Running command ``` docker run -v $(pwd)/data/:/home/isr/data -v $(pwd)/weights/:/home/isr/weights -v $(pwd)/config.yml:/home/isr/config.yml -it isr -p -d -c config.yml ``` Gets error message ```Traceback (most recent call last): File "/home/isr/ISR/assistant.py", line 90,...
Dear cfrancesco, How can I train, with initial weights? As you suggested, I want to choose psnr loss for 50+ first epochs, then switch to gans. So when I start...
The issue caused the colab notebooks to not work and fail. This fixes the conflict issue with the package `h5py`. **Thank you for this amazing repository. It really has helped...
When trying to use predictor, I cannot load weights: ```{python} --------------------------------------------------------------------------- AttributeError Traceback (most recent call last) in 2 #!mkdir -p data/output 3 predictor = Predictor(input_dir='test_data/input/') ----> 4 predictor.get_predictions(model=rrdn, weights_path='weights/rrdn-C4-D3-G32-G032-T10-x4/2021-03-05_1158/rrdn-C4-D3-G32-G032-T10-x4_epoch009.hdf5')...
Wondering what loss functions the RDN model was trained on. For the 'noise-cancel' weights, it says that it was trained with adversarial and VGG feature losses - is this also...
I attempted to use TIFF raster data, first raw and then converted to jpg and png. Every time it returns a shape error. Is there something I can do to...
I installed the library, ISR. When I run the code, from ISR.models import RDN, RRDN the error, ModuleNotFoundError: No module named 'wrapt', popped up. it seems there are some version...
