css change

app.py

@@ -13,7 +13,7 @@ from timm.data import resolve_data_config
 from timm.data.transforms_factory import create_transform
 
 from model import Model
-from get_output import visualize_output
+from output import visualize_output
 
 
 # Use GPU if available
@@ -30,13 +30,13 @@ model.eval()
 state = torch.load('saved_model', map_location=torch.device('cpu'))
 model.load_state_dict(state['val_model_dict'])
 
-# Transform for input image
+# Create transform for input image
 config = resolve_data_config({}, model=vit)
 config['no_aug'] = True
 config['interpolation'] = 'bilinear'
 transform = create_transform(**config)
 
-
+# Inference function
 def query_image(input_img, query, binarize, eval_threshold):
 
     PIL_image = Image.fromarray(input_img, "RGB")
@@ -49,10 +49,10 @@ def query_image(input_img, query, binarize, eval_threshold):
     img = visualize_output(img, output, binarize, eval_threshold)
     return img
 
-
+# Gradio interface
 description = """
 Gradio demo for an object detection architecture, 
-introduced in <a href="https://arxiv.org/abs/2205.06230">my bachelor thesis</a>. 
+introduced in <a href="https://www.google.com/">my bachelor thesis (link will be added)</a>. 
 \n\nLorem ipsum ....
 *"image of a shoe"*. Refer to the <a href="https://arxiv.org/abs/2103.00020">CLIP</a> paper to see the full list of text templates used to augment the training data.
 """
@@ -67,6 +67,9 @@ demo = gr.Interface(
     ],
     allow_flagging = "never",
     cache_examples=False,
+    css = """
+    body {background-color : grey}
+    """,
 )
 demo.launch(debug=True)
 

model.py

@@ -1,7 +1,6 @@
 import torch
 import torch.nn as nn
 
-
 class Model(nn.Module):
     def __init__(self, vit, roberta, tokenizer, device):
         super().__init__()

get_output.py → output.py

@@ -25,6 +25,7 @@ def enlarge_array(output):
     
     return output
 
+
 def visualize_output(image, output, binarize, threshold):
     
     image, output = preprocess(image, output, binarize, threshold) 
@@ -35,9 +36,9 @@ def visualize_output(image, output, binarize, threshold):
     plt.axis('off')
     plt.imshow(image)
     if binarize:
-        plt.imshow(output_mask, alpha=.67)
+        plt.imshow(output_mask, alpha=.5)
     else:
-        plt.imshow(output_mask, alpha=.8)
+        plt.imshow(output_mask, alpha=.6)
     fig.tight_layout(pad=0)
     fig.canvas.draw()
     data = np.frombuffer(fig.canvas.tostring_rgb(), dtype=np.uint8)