Update app.py

app.py

@@ -7,7 +7,7 @@ from torch.nn.functional import cosine_similarity
 import gradio as gr
 
 class RoiMatching():
-    def __init__(self,img1,img2,device='cuda:1', v_min=200, v_max= 7000, mode = 'embedding'):
+    def __init__(self,img1,img2,device='cpu', v_min=200, v_max= 7000, mode = 'embedding'):
         """
         Initialize
         :param img1: PIL image
@@ -230,11 +230,11 @@ def predict(im1,im2):
     visualized_image1, visualized_image2 = visualize_masks(im1, RM.masks1, im2, RM.masks2)
     return visualized_image1, visualized_image2
 
-# examples = [
-            # ['./example/prostate_2d/image1.png', './example/prostate_2d/image2.png'],
-            # ['./example/cardiac_2d/image1.png', './example/cardiac_2d/image2.png'],
+examples = [
+            ['./example/prostate_2d/image1.png', './example/prostate_2d/image2.png'],
+            ['./example/cardiac_2d/image1.png', './example/cardiac_2d/image2.png'],
             # ['./example/pathology/1B_B7_R.png', './example/pathology/1B_B7_T.png'],
-           # ]
+           ]
 
 
 gradio_app = gr.Interface(
@@ -242,7 +242,7 @@ gradio_app = gr.Interface(
     inputs=[gr.Image(label="img1", type="pil"), gr.Image(label="img2", type="pil")],
     outputs=[gr.Image(label="ROIs in img1"), gr.Image(label="ROIs in img2")],
     title="SAMReg: One Registration is Worth Two Segmentations",
-    # examples=examples,
+    examples=examples,
     description="<p> \
                     <strong>Register anything with ROI-based registration representation.</strong> <br>\
                     Choose an example below &#128293; &#128293;  &#128293; <br>\
@@ -253,4 +253,6 @@ gradio_app = gr.Interface(
                             💎 Examples below are all medical images for the algorithm proposed for medical registration initially. <br>\
                             💎 Current UI interface only unleashes a small part of the capabilities of SAMReg, i.e., 2D registration w 'embedding' mode. \
 </p>",
+    cache_examples=False,
+    allow_flagging="never",
 )