Update app.py

app.py

@@ -7,79 +7,99 @@ from PIL import Image
 from pytorch_grad_cam import GradCAM
 from pytorch_grad_cam.utils.image import show_cam_on_image
 from resnetS11 import LITResNet
+import os
+import re
+import matplotlib.pyplot as plt
+from io import BytesIO
 
 inv_normalize = transforms.Normalize(
     mean=[-0.50/0.23, -0.50/0.23, -0.50/0.23],
     std=[1/0.23, 1/0.23, 1/0.23]
 )
-
-classes = ('plane', 'car', 'bird', 'cat', 'deer',
-           'dog', 'frog', 'horse', 'ship', 'truck')
+classes = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
 
 model = LITResNet(classes)
 model.load_state_dict(torch.load("model.pth", map_location=torch.device('cpu')), strict=False)
 
-def inference(input_img, num_gradcam_images=1, target_layer_number=-1, transparency=0.5, show_misclassified=False, num_top_classes=3):
-    input_img = np.array(Image.fromarray(np.array(input_img)).resize((32,32)))
+def inference(input_img, num_gradcam_images=1, target_layer_number=-1, transparency=0.5, show_misclassified=False, num_top_classes=3, num_misclassified_images=3):
+    input_img = np.array(Image.fromarray(np.array(input_img)).resize((32, 32)))
     org_img = input_img
-
     transform = transforms.ToTensor()
     input_img = transform(input_img).unsqueeze(0)
-
     outputs = model(input_img)
     softmax = torch.nn.Softmax(dim=0)
     o = softmax(outputs.flatten())
     confidences = {classes[i]: float(o[i]) for i in range(10)}
     _, prediction = torch.max(outputs, 1)
 
-    if not show_misclassified or prediction[0].item() == np.argmax(list(confidences.values())):
-        target_layers = [model.layer2[target_layer_number]]
-        cam = GradCAM(model=model, target_layers=target_layers)
-        grayscale_cam = cam(input_tensor=input_img, targets=None)
-        grayscale_cam = grayscale_cam[0, :]
+    target_layers = [[model.X3], [model.R3]]
+    cam = GradCAM(model=model, target_layers=target_layers[target_layer_number], use_cuda=False)
+    grayscale_cam = cam(input_tensor=input_img, targets=None)
+    grayscale_cam = grayscale_cam[0, :]
+    img = input_img.squeeze(0)
+    img = inv_normalize(img)
+    rgb_img = np.transpose(img, (1, 2, 0))
+    rgb_img = rgb_img.numpy()
+    visualization = show_cam_on_image(org_img/255, grayscale_cam, use_rgb=True, image_weight=transparency)
+
+    # Sort the confidences dictionary based on confidence values
+    sorted_confidences = dict(sorted(confidences.items(), key=lambda item: item[1], reverse=True))
+
+    # Pick the top n predictions
+    top_n_confidences = dict(list(sorted_confidences.items())[:num_top_classes])
 
-        img = input_img.squeeze(0)
-        img = inv_normalize(img)
+    if show_misclassified:
+        files = os.listdir('./misclassfied_images/')
 
-        visualizations = []
-        for _ in range(num_gradcam_images):
-            visualization = show_cam_on_image(org_img/255, grayscale_cam, use_rgb=True, image_weight=transparency)
-            visualizations.append(visualization)
+        # Plot the misclassified images
+        fig = plt.figure(figsize=(12, 5))
+        for i in range(num_misclassified_images):
+            sub = fig.add_subplot(2, 5, i+1)
+            npimg = Image.open('./misclassfied_images/' + files[i])
 
-        top_classes = sorted(confidences.items(), key=lambda x: x[1], reverse=True)[:num_top_classes]
-        top_classes_dict = {cls: conf for cls, conf in top_classes}
+            # Use regex to extract target and predicted classes
+            match = re.search(r'Target_(\w+)_Pred_(\w+)_\d+.jpeg', files[i])
+            target_class = match.group(1)
+            predicted_class = match.group(2)
 
-        return prediction[0].item(), visualizations, top_classes_dict
+            plt.imshow(npimg, cmap='gray', interpolation='none')
+            sub.set_title("Actual: {}, Pred: {}".format(target_class, predicted_class), color='red')
+        plt.tight_layout()
+        buffer = BytesIO()
+        plt.savefig(buffer, format='png')
+        visualization_misclassified = Image.open(buffer)
+
+        return top_n_confidences, visualization, visualization_misclassified
     else:
-        return None, None, None
+        return top_n_confidences, visualization
 
 title = "CIFAR10 trained on ResNet18 Model with GradCAM"
 description = "A simple Gradio interface to infer on ResNet model, and get GradCAM results"
-
-examples = [
-    ["cat.jpg", 1, -1, 0.5, False, 3],
-    ["dog.jpg", 1, -1, 0.5, False, 3],
-    # Add more example images here
-]
-
+examples = [["cat.jpg", 1, -1, 0.8, True, 3, 3],
+            ["dog.jpg", 1, -1, 0.8, True, 3, 3],
+            # ["plane.jpeg", 1, -1, 0.8, True, 3, 3],
+            # ["deer.jpeg", 1, -1, 0.8, True, 3, 3],
+            # ["horse.jpeg", 1, -1, 0.8, True, 3, 3],
+            # ["bird.jpeg", 1, -1, 0.8, True, 3, 3],
+            # ["frog.jpeg", 1, -1, 0.8, True, 3, 3],
+            # ["ship.jpeg", 1, -1, 0.8, True, 3, 3],
+            # ["truck.jpeg", 1, -1, 0.8, True, 3, 3],
+            # ["car.jpeg", 1, -1, 0.8, True, 3, 3]]
+           ]
 demo = gr.Interface(
     inference,
-    inputs=[
-        gr.Image(width=256, height=256, label="Input Image"),
-        gr.Slider(1, 5, value=1, step=1, label="Number of GradCAM Images"),
-        gr.Slider(-2, -1, value=-2, step=1, label="Which Layer?"),
-        gr.Slider(0, 1, value=0.5, label="Overall Opacity of Image"),
-        gr.Checkbox(label="Show Misclassified Images"),
-        gr.Slider(1, 10, value=3, step=1, label="Number of Top Classes")
-    ],
-    outputs=[
-        "text",
-        gr.Gallery(label="GradCAM Images"),
-        gr.Label(num_top_classes=3, label="Top Classes")
-    ],
+    inputs=[gr.Image(shape=(32, 32), label="Input Image"),
+            gr.Slider(1, 10, value=1, step=1, label="Number of GradCAM Images"),
+            gr.Slider(-2, -1, value=-1, step=1, label="Which Layer?"),
+            gr.Slider(0, 1, value=0.8, label="Opacity of GradCAM"),
+            gr.Checkbox(value=True, label="Show Misclassified Images"),
+            gr.Slider(2, 10, value=3, step=1, label="Top Predictions"),
+            gr.Slider(1, 10, value=3, step=1, label="Misclassified Images")],
+    outputs=[gr.Label(label="Top Predictions"),
+             gr.Image(shape=(32, 32), label="Output").style(width=128, height=128),
+             gr.Image(shape=(640, 360), label="Misclassified Images").style(width=640, height=360)],
     title=title,
     description=description,
     examples=examples,
 )
-
 demo.launch()