app fixed for description and aspect ratio

app.py

@@ -72,7 +72,9 @@ def predict(img):
     """
     image_size = (256,256)
     upscale_factor = 4
-    lr_transforms = transforms.Resize((image_size[0]//upscale_factor, image_size[1]//upscale_factor), interpolation=IMode.BICUBIC, antialias=True)
+    # lr_transforms = transforms.Resize((image_size[0]//upscale_factor, image_size[1]//upscale_factor), interpolation=IMode.BICUBIC, antialias=True)
+    # to retain aspect ratio
+    lr_transforms = transforms.Resize(image_size[0]//upscale_factor, interpolation=IMode.BICUBIC, antialias=True)
     # lr_transforms = transforms.Resize((128, 128), interpolation=IMode.BICUBIC, antialias=True)
     
     img = Image.fromarray(np.array(img))
@@ -114,7 +116,8 @@ def predict(img):
 import gradio as gr
 
 title = "BayesCap"
-description = "BayesCap: Bayesian Identity Cap for Calibrated Uncertainty in Frozen Neural Networks (ECCV 2022)"
+method = "In this work, we propose a method (called BayesCap) to estimate the per-pixel uncertainty of a pretrained computer vision model like SRGAN (used for super-resolution). \n BayesCap takes the ouput of the pretrained model (in this case SRGAN), and predicts the per-pixel distribution parameters for the output, that can be used to quantify the per-pixel uncertainty. In our work, we model the per-pixel output as a <a href='https://en.wikipedia.org/wiki/Generalized_normal_distribution'>Generalized Gaussian distribution</a> that is parameterized by 3 parameters the mean, scale (alpha), and the shape (beta). As a result our model predicts these three parameters as shown below. From these 3 parameters one can compute the uncertainty as shown in <a href='https://en.wikipedia.org/wiki/Generalized_normal_distribution'>this article</a>."
+description = "BayesCap: Bayesian Identity Cap for Calibrated Uncertainty in Frozen Neural Networks (ECCV 2022) \n" + method
 article = "<p style='text-align: center'> BayesCap: Bayesian Identity Cap for Calibrated Uncertainty in Frozen Neural Networks| <a href='https://github.com/ExplainableML/BayesCap'>Github Repo</a></p>"