improved ui?

.vscode/settings.json

@@ -0,0 +1,3 @@
+{
+    "python-envs.defaultEnvManager": "ms-python.python:pipenv"
+}

TestSampleDataset.py

@@ -0,0 +1,27 @@
+import numpy as np
+import torch
+from torch.utils.data import Dataset
+
+
+class TestDataset(Dataset):
+
+    def __init__(self, npz_path, transform=None):
+        data = np.load(npz_path)
+
+        self.images = data["images"]
+        self.labels = data["labels"]
+        self.transform = transform
+
+    def __len__(self):
+        return len(self.images)
+
+    def __getitem__(self, idx):
+        image = self.images[idx]
+        label = self.labels[idx]
+
+        if self.transform:
+            image = self.transform(image)
+
+        label = torch.tensor(label, dtype=torch.float32).squeeze()
+
+        return image, label

app.py

@@ -5,19 +5,23 @@ import gradio
 from PIL import Image
 from huggingface_hub import hf_hub_download
 from models.linear_predictor import Predictor
+from torch.utils.data import Dataset
+from TestSampleDataset import TestDataset
+import numpy
+import os
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
 LABELS = [
-    "adipose",
-    "background",
-    "debris",
-    "lymphocytes",
-    "mucus",
-    "smooth muscle",
-    "normal colon mucosa",
-    "cancer-associated stroma",
-    "colorectal adenocarcinoma epithelium",
+    "Adipose",
+    "Background",
+    "Debris",
+    "Lymphocytes",
+    "Mucus",
+    "Smooth Muscle",
+    "Normal Colon Mucosa",
+    "Cancer-associated Stroma",
+    "Colorectal Adenocarcinoma Epithelium",
 ]
 
 
@@ -32,19 +36,37 @@ model.load_state_dict(torch.load(model_file,map_location=device))
 model.to(device)
 model.eval()
 
-
 tf = v2.Compose([
     v2.ToImage(),
     v2.Resize((64, 64)),
     v2.ToDtype(torch.float32, scale=True),
 ])
 
+dataset = numpy.load("test_samples.npz")
+images = dataset["images"]
+labels = dataset["labels"]
+
+example_rows = []
+number_of_examples = len(labels)
+
+os.makedirs("ui_examples", exist_ok=True)
+
+for i in range(number_of_examples):
+    img_array = images[i]
+    label_index = int(labels[i].item() if hasattr(labels[i], 'item') else labels[i])
+    
+    truth_label_text = LABELS[label_index] if label_index < len(LABELS) else f"Class {label_index}"
+    
+    file_path = f"ui_examples/sample_{i}.jpg"
+    Image.fromarray(img_array.astype("uint8"), "RGB").save(file_path)
+
+    example_rows.append([file_path, truth_label_text])
+
 @spaces.GPU
 def predict(image):
     if image is None:
         return None
     
-    
     pil_img = Image.fromarray(image.astype('uint8'), 'RGB')
     
     img_tensor = tf(pil_img).unsqueeze(0).to(device)
@@ -55,13 +77,29 @@ def predict(image):
     
     return {LABELS[i]: float(probabilities[i]) for i in range(len(LABELS))}
 
-demo = gradio.Interface(
-    fn=predict,
-    inputs=gradio.Image(), 
-    outputs=gradio.Label(num_top_classes=3),
-    title="PathMNIST Image Classification",
-    description="Upload a tissue patch image for classification"
-)
+with gradio.Blocks() as demo:
+    gradio.Markdown("# PathMNIST Image Classification")
+
+    with gradio.Tab("Predict"):
+        gradio.Markdown("Upload a tissue patch image for classification.")
+        input_img = gradio.Image()
+        output_lbl = gradio.Label(num_top_classes=3)
+        btn = gradio.Button("Predict")
+        btn.click(fn=predict, inputs=input_img, outputs=output_lbl)
+
+    with gradio.Tab("Examples"):
+        gradio.Markdown("Click an example below to test the model against the PathMNIST test dataset.")
+        
+        # Create a read-only text box to display the column for Truth Labels
+        truth_box = gradio.Textbox(label="Ground Truth Label", interactive=False)
+        
+        gradio.Examples(
+            examples=example_rows,  # Passes both the image file path and the text label
+            inputs=[input_img, truth_box],  # Maps the data columns to both UI components
+            outputs=output_lbl,
+            fn=predict,
+            cache_examples=True,
+        )
 
 if __name__ == "__main__":
     demo.launch()

requirements.txt

@@ -1,6 +1,7 @@
 gradio==6.24.0
 huggingface_hub
 torch
+numpy
 torchvision
 pillow
 git+https://github.com/HasanAli5/MAE-Model-MedMNIST-Predictor.git#egg=mae_model

test_samples.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36bc98ea602b8a622548839906cbb2aaf67ab8cc17e78e94a1a08b0e342ede4a
+size 2548100