Basic version is complete and working.

app.py

@@ -1,17 +1,25 @@
 import gradio as gr
-from transformers import pipeline
+import torch
+from torchvision.transforms.functional import pil_to_tensor, to_pil_image
 
-pipeline = pipeline(task="image-classification", model="julien-c/hotdog-not-hotdog")
+model = torch.jit.load("models/lama.pt")
 
-def predict(input_img):
-    predictions = pipeline(input_img)
-    return input_img, {p["label"]: p["score"] for p in predictions}
+def predict(input_img, input_mask):
+    # numpy gives the image as (w,h,c)
+    # Image shape should be (1, 3, 512, 512) and be in the range 0-1.
+    # Mask shape should be (1, 1, 512, 512) AND have values 0.0 or 1.0, not in-between.
+    #out = model(torch.tensor(input_img[None, (2,0,1), :, :])/255.0, torch.tensor(1 * (input_mask[:,:,0] > 0)).unsqueeze(0))
+    out = model((pil_to_tensor(input_img.convert('RGB')) / 255.0).unsqueeze(0), 1 * (pil_to_tensor(input_mask.convert('L')) > 0).unsqueeze(0))[0]
+    return to_pil_image(out)
 
 gradio_app = gr.Interface(
     predict,
-    inputs=gr.Image(label="Select hot dog candidate", sources=['upload', 'webcam'], type="pil"),
-    outputs=[gr.Image(label="Processed Image"), gr.Label(label="Result", num_top_classes=2)],
-    title="Hot Dog? Or Not?",
+    inputs=[
+        gr.Image(label="Select Base Image", sources=['upload',], type="pil"), 
+        gr.Image(label="Select Image Mask (White will be inpainted)", sources=['upload',], type="pil"), 
+    ],
+    outputs=[gr.Image(label="Inpainted Image"),],
+    title="LAMA Inpainting",
 )
 
 if __name__ == "__main__":

pickle_lama_model.ipynb

@@ -160,7 +160,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 69,
+   "execution_count": 76,
    "id": "163db07c-93a3-40d2-837d-4fade79b07f0",
    "metadata": {},
    "outputs": [
@@ -181,12 +181,22 @@
      },
      "metadata": {},
      "output_type": "display_data"
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "tensor(1.)\n",
+      "tensor(1)\n"
+     ]
     }
    ],
    "source": [
     "print(out['predicted_image'].shape)\n",
     "import numpy\n",
-    "display(tvf.to_pil_image((out['predicted_image'])[0]))"
+    "display(tvf.to_pil_image((out['predicted_image'])[0]))\n",
+    "print(torch.max(image))\n",
+    "print(torch.max(mask))"
    ]
   },
   {