change image to pil format

app.py

@@ -94,7 +94,7 @@ class Network(nn.Module):
 model = load_checkpoint("flower_inference_model.pth")
 
 
-def process_image(img_path):
+def process_image(pil_image):
     """Scales, crops, and normalizes a PIL image for a PyTorch model,
     returns a Numpy array
 
@@ -102,7 +102,7 @@ def process_image(img_path):
     ---------
     image: path of the image to be processed
     """
-    inp = Image.open(img_path)
+    inp = Image.open(pil_image)
     inp.load()
     img_exif = inp.getexif()
 
@@ -117,12 +117,12 @@ def process_image(img_path):
     image_path = re.sub(r"\W+", "_", image_path)
 
     # Join to directory path
-    inf_image = os.path.join("inference", image_path)
+    inf_image = os.path.join("inference", image_path + ".jpg")
 
     # Use repo for inference
-    inp.save(inf_image, format="JPEG", quality=95, keep=True, exif=img_exif)
+    inp.save(inf_image, quality=95, keep=True, exif=img_exif)
     HfApi().upload_file(
-        path_or_fileobj=inf_image + ".JPG",
+        path_or_fileobj=inf_image,
         path_in_repo=image_path,
         repo_id="DanielPFlorian/flower-image-classifier",
         repo_type="dataset",
@@ -164,7 +164,7 @@ with open("cat_to_name.json", "r") as f:
     cat_to_name = json.load(f)
 
 
-def predict(image_path, model=model, category_names=cat_to_name, topk=5):
+def predict(pil_image, model=model, category_names=cat_to_name, topk=5):
     """Predict the class (or classes) of an image using a trained deep learning model.
     Arguments
     ---------
@@ -173,7 +173,7 @@ def predict(image_path, model=model, category_names=cat_to_name, topk=5):
     topk: number of top predicted classes to return
     """
     # Process image function
-    image = process_image(image_path)
+    image = process_image(pil_image)
 
     # Convert image to float tensor with batch size of 1
     image = torch.as_tensor(image).view((1, 3, 224, 224)).float()
@@ -210,7 +210,7 @@ def predict(image_path, model=model, category_names=cat_to_name, topk=5):
 
         # Plot Functionality
 
-        image = Image.open(image_path)
+        image = Image.open(pil_image)
         fig, (ax1, ax2) = plt.subplots(ncols=2)
         ax1.imshow(image)
         ax1.axis("off")
@@ -231,7 +231,7 @@ def predict(image_path, model=model, category_names=cat_to_name, topk=5):
 # Gradio Interface
 gr.Interface(
     predict,
-    inputs=gr.inputs.Image(label="Upload a flower image", type="filepath"),
+    inputs=gr.inputs.Image(label="Upload a flower image", type="pil"),
     outputs=gr.Plot(label="Plot"),
     title="What kind of flower is this?",
 ).launch()