Update app.py

app.py

@@ -10,8 +10,15 @@ from pathlib import Path
 import random
 
 import torchvision.transforms as transforms
+import PIL
+
+import gradio as gr
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
 model = load_learner('export (2).pkl')
+model.cpu()
+model.eval()
 
 def transform_image(image):
    my_transforms = transforms.Compose([transforms.ToTensor(), transforms.Normalzie([0.485, 0.456, 0.406],[0.229, 0.224, 0.225])])
@@ -19,12 +26,14 @@ def transform_image(image):
 
 def predict(img):
    img = PILImage.create(img)
-   device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+   
    image = transforms.Resize((480,640))(img)
    tensor = transform_image(image=image)
-   model.to(device)
    with torch.no_grad():
       outputs = model(tensor)
+   
+   outputs = torch.argmax(outputs,1)
+   
    mask = np.array(outputs.cpu())
    mask[mask==0]=255 
    mask[mask==1]=150
@@ -32,6 +41,7 @@ def predict(img):
    mask[mask==3]=25
    mask[mask==4]=0
    mask=np.reshape(mask,(480,640))
-   Image.fromarray(mask.astype('uint8')) 
    
-gr.Interface(fn=predict, inputs=gr.inputs.Image(shape=(128,128)), outputs=gr.inputs.Image(shape=(128,128)), examples=['color_157.jpg','color_158.jpg']).launch(share=False)
+   return Image.fromarray(mask.astype('uint8'))
+   
+gr.Interface(fn=predict, inputs=gr.inputs.Image(shape=(128,128)), outputs=gr.inputs.Image(), examples=['color_157.jpg','color_158.jpg']).launch(share=False)