add app.py

app.py

@@ -0,0 +1,125 @@
+
+import torch, torchvision
+from monai.networks.nets import UNet
+from monai.networks.layers import Norm
+from monai.inferers import sliding_window_inference
+import PIL
+from torchvision.utils import save_image
+import numpy as np
+
+model = UNet(
+    spatial_dims=3,
+    in_channels=1,
+    out_channels=2,
+    channels=(16, 32, 64, 128, 256),
+    strides=(2, 2, 2, 2),
+    num_res_units=2,
+    norm=Norm.BATCH,
+)
+
+model.load_state_dict(torch.load("weights/model.pt", map_location=torch.device('cpu')))
+
+import gradio as gr
+
+def load_image0():
+    return load_image(0)
+
+def load_image1():
+    return load_image(1)
+
+def load_image2():
+    return load_image(2)
+
+def load_image3():
+    return load_image(3)
+
+def load_image4():
+    return load_image(4)
+
+def load_image5():
+    return load_image(5)
+
+def load_image6():
+    return load_image(6)
+
+def load_image7():
+    return load_image(7)
+
+def load_image8():
+    return load_image(8)
+
+def load_image(index):
+    return [index, f"thumbnails/val_image{index}.png", f"thumbnails_label/val_label{index}.png"]
+
+def predict(index):
+    val_data = torch.load(f"samples/val_data{index}.pt")
+    
+    model.eval()
+    with torch.no_grad():
+        roi_size = (160, 160, 160)
+        sw_batch_size = 4
+        val_outputs = sliding_window_inference(val_data, roi_size, sw_batch_size, model)
+
+    meta_tsr = torch.argmax(val_outputs, dim=1)[0, :, :, 80]
+    pil_image = torchvision.transforms.functional.to_pil_image(meta_tsr.to(torch.float32))
+
+    return pil_image
+
+
+with gr.Blocks(title="Spleen 3D segmentation with MONAI - ClassCat",
+                    css=".gradio-container {background:azure;}"
+               ) as demo:
+    sample_index = gr.State([])
+
+    gr.HTML("""<div style="font-family:'Times New Roman', 'Serif'; font-size:16pt; font-weight:bold; text-align:center; color:royalblue;">Spleen 3D segmentation with MONAI</div>""")
+
+    gr.HTML("""<h4 style="color:navy;">1. Select an example, which includes input images and label images, by clicking "Example x" button.</h4>""")
+
+    with gr.Row():
+        input_image = gr.Image(label="a piece of input image data", type="filepath")
+        label_image = gr.Image(label="label image", type="filepath")
+        output_image = gr.Image(label="predicted image", type="pil")
+
+
+    with gr.Row():
+        with gr.Column():
+            ex_btn0 = gr.Button("Example 1")
+            ex_btn0.style(full_width=False, css="width:20px;")
+            ex_image0 = gr.Image(value='thumbnails/val_image0.png', interactive=False, label='ex 1')
+            ex_image0.style(width=128, height=128)
+
+        with gr.Column():
+            ex_btn1 = gr.Button("Example 2")
+            ex_btn1.style(full_width=False, css="width:20px;")
+            ex_image1 = gr.Image(value='thumbnails/val_image1.png', interactive=False, label='ex 2')
+            ex_image1.style(width=128, height=128)
+
+        with gr.Column():
+            ex_btn2 = gr.Button("Example 3")
+            ex_btn2.style(full_width=False, css="width:20px;")
+            ex_image2 = gr.Image(value='thumbnails/val_image2.png', interactive=False, label='ex 3')
+            ex_image2.style(width=128, height=128)
+
+        with gr.Column():
+            ex_btn3 = gr.Button("Example 4")
+            ex_btn3.style(full_width=False, css="width:20px;")
+            ex_image3 = gr.Image(value='thumbnails/val_image3.png', interactive=False, label='ex 4')
+            ex_image3.style(width=128, height=128)
+
+    ex_btn0.click(fn=load_image0, outputs=[sample_index, input_image, label_image])
+    ex_btn1.click(fn=load_image1, outputs=[sample_index, input_image, label_image])
+    ex_btn2.click(fn=load_image2, outputs=[sample_index, input_image, label_image])
+    ex_btn3.click(fn=load_image3, outputs=[sample_index, input_image, label_image])
+
+    gr.HTML("""<br/>""")
+    gr.HTML("""<h4 style="color:navy;">2. Then, click "Infer" button to predict segmentation images. It will take about 30 seconds (on cpu)</h4>""")
+
+    send_btn = gr.Button("Infer")
+    send_btn.click(fn=predict, inputs=[sample_index], outputs=[output_image])
+
+
+#demo.queue()
+demo.launch(debug=True)
+
+
+### EOF ###