Fixed the interface

app.py

@@ -1,8 +1,9 @@
 import gradio as gr
 from fastai.vision.all import *
+from tools import *
 import skimage
 
-learn = load_learner('models/panda-model-1.pth')
+learn = load_learner('panda-model-1.pkl')
 
 labels = learn.dls.vocab
 
@@ -14,8 +15,6 @@ def predict(img):
 title = "Prostate cANcer graDe Assessment model"
 description = "A model to predict the ISUP grade for prostate cancer based on whole-slide images of digitized H&E-stained biopsies."
 # article="<p style='text-align: center'><a href='https://tmabraham.github.io/blog/gradio_hf_spaces_tutorial' target='_blank'>Blog post</a></p>"
-examples = ['test.jpg']
-interpretation='default'
-enable_queue=True
+examples = ['example.jpg', 'example2.jpg', 'example3.jpg']
 
-gr.Interface(fn=predict,inputs=gr.inputs.Image(shape=(224, 224)),outputs=gr.outputs.Label(num_top_classes=3),title=title,description=description,article=article,examples=examples,interpretation=interpretation,enable_queue=enable_queue).launch()
+gr.Interface(fn=predict,inputs=gr.inputs.Image(),outputs=gr.outputs.Label(num_top_classes=5),title=title,description=description,examples=examples).launch()

tools.py

@@ -10,13 +10,15 @@ def get_crops(x):
     else:
         tiff_file = imgdir/f'{x["image_id"]}.tiff'
         img = tifffile.imread(tiff_file, key=1)
+    print('input image shape:', img.shape)
     crop = np.array(img.shape) // tile_size * tile_size; crop
     imgc = img[:crop[0],:crop[1]]
     imgc = imgc.reshape(imgc.shape[0] // tile_size, tile_size, imgc.shape[1] // tile_size, tile_size, 3)
     xs, ys = (imgc.mean(axis=1).mean(axis=2).mean(axis=-1) < 252).nonzero()
     if len(xs) == 0:
         xs, ys = (imgc.mean(axis=1).mean(axis=2).mean(axis=-1)).nonzero()
-#     if len(xs) < 2: print("no data in image:", x)
+    if len(xs) < 2: print("no data in image:", x)
+    print('len xs:', len(xs))
     pidxs = random.choices(list(range(len(xs))), k=36)
     return PILImage.create(imgc[xs[pidxs],:,ys[pidxs],:].reshape(6,6,tile_size,tile_size,3).transpose(0,2,1,3,4).reshape(6*tile_size,6*tile_size,3))
 #     return imgc.mean(axis=1).mean(axis=2).mean(axis=-1)