Update app.py

app.py

@@ -6,6 +6,7 @@ from torchvision import models
 from torch import nn 
 from typing import List
 import json
+import pandas as pd
 
 #Read labels file1
 with open('cat_to_name.json','r') as f:
@@ -69,37 +70,54 @@ def pred_image(model, image_path, class_names = None, transform=None, device: to
 def process_input(image_path):
 
     #Load Model
-    model_name, model_weights, model_path = ('efficientnet_b2','EfficientNet_B2_Weights','flowers_efficientnet_b2_model.pth')
+    list_of_models_and_weights = [
+    ('efficientnet_b2','EfficientNet_B2_Weights','flowers_efficientnet_b2_model.pth'),
+    ('alexnet','AlexNet_Weights','flowers_alexnet_model.pth')   
+    #('mobilenet_v2','MobileNet_V2_Weights','flowers_mobilenet_v2_model.pth'),
+    #('densenet121','DenseNet121_Weights','flowers_densenet121_model.pth'),
+    #('inception_v3','Inception_V3_Weights','flowers_inception_v3_model.pth'),
+    #('squeezenet1_1','SqueezeNet1_1_Weights','flowers_squeezenet1_1_model.pth'),
+    #('vgg16','VGG16_Weights','flowers_vgg16_model.pth'),
+    #('resnet18','ResNet18_Weights','flowers_resnet18_model.pth'),
+    #('swin_b','Swin_B_Weights','flowers_swin_b_model.pth'),
+    #('vit_b_16', 'ViT_B_16_Weights','flowers_vit_b_16_model.pth')
+                                   ] 
+    #model_name, model_weights, model_path = ('efficientnet_b2','EfficientNet_B2_Weights','flowers_efficientnet_b2_model.pth')
     #model_name, model_weights, model_path = ('alexnet','AlexNet_Weights','flowers_alexnet_model.pth')
-    checkpoint = torch.load(model_path, map_location='cpu')
-    pretrained_weights = eval(f"models.{model_weights}.DEFAULT")
-    auto_transforms = pretrained_weights.transforms()
-    #pretrained_model = eval(f"torchvision.models.{model_name}(weights = pretrained_weights)")
-    pretrained_model = eval(f"models.{model_name}(pretrained = True)")
-    pretrained_model = update_last_layer_pretrained_model(pretrained_model, 102, True)
-    pretrained_model.class_to_idx = checkpoint['class_to_idx']
-    pretrained_model.class_names = checkpoint['class_names']    
-    pretrained_model.load_state_dict(checkpoint['state_dict'])
-    pretrained_model.to('cpu')
+    list_of_outputs = []
+    for model_name, model_weights, model_path in list_of_models_and_weights:
+        checkpoint = torch.load(model_path, map_location='cpu')
+        pretrained_weights = eval(f"models.{model_weights}.DEFAULT")
+        auto_transforms = pretrained_weights.transforms()
+        #pretrained_model = eval(f"torchvision.models.{model_name}(weights = pretrained_weights)")
+        pretrained_model = eval(f"models.{model_name}(pretrained = True)")
+        pretrained_model = update_last_layer_pretrained_model(pretrained_model, 102, True)
+        pretrained_model.class_to_idx = checkpoint['class_to_idx']
+        pretrained_model.class_names = checkpoint['class_names']    
+        pretrained_model.load_state_dict(checkpoint['state_dict'])
+        pretrained_model.to('cpu')
+            
+        #Predict
+        #image_path = 'which-flower/80_image_02020.jpg'
+        probs, idxs = pred_image(model=pretrained_model, image_path=image_path, class_names=pretrained_model.class_names, transform=auto_transforms)
+        names = [cat_to_name[i] for i in idxs]
         
-    #Predict
-    #image_path = 'which-flower/80_image_02020.jpg'
-    probs, idxs = pred_image(model=pretrained_model, image_path=image_path, class_names=pretrained_model.class_names, transform=auto_transforms)
-    names = [cat_to_name[i] for i in idxs]
-    
+        list_of_outputs.append({"Prediction" : names[0], "Probability" : probs[0]})
+            
     #Display or return to main function
-    print({names[i]: float(probs[i]) for i in range(len(names))})
-    return {names[i]: float(probs[i]) for i in range(len(names))}, {names[i]: float(probs[i]) for i in range(len(names))}
+    #print({names[i]: float(probs[i]) for i in range(len(names))})
     #return {names[i]: float(probs[i]) for i in range(len(names))}
-                    
+    #oldreturn {names[i]: float(probs[i]) for i in range(len(names))}
+    print(pd.DataFrame(list_of_outputs))
+    return pd.DataFrame(list_of_outputs)
+                        
 examples = ['16_image_06670.jpg','33_image_06460.jpg','80_image_02020.jpg', 'Flowers.png','inference_example.png']
 title = "Image Classifier - Species of Flower predicted by different Models"
 description = "Image classifiers to recognize different species of flowers trained on 102 Category Flower Dataset"
 article = article="<p style='text-align: center'><a href='https://www.robots.ox.ac.uk/~vgg/data/flowers/102/index.html' target='_blank'>Source 102 Flower Dataset</a></p>"
 interpretation = 'default'
 enable_queue = True
-iface = gr.Interface(fn=process_input, inputs=gr.inputs.Image(type='filepath'), outputs=[gr.outputs.Label(num_top_classes=3), gr.outputs.Label(num_top_classes=3)], examples = examples, title=title, description=description,article=article,interpretation=interpretation, enable_queue=enable_queue
-                    )
+iface = gr.Interface(fn=process_input, inputs=gr.inputs.Image(type='filepath'), outputs="dataframe", examples = examples, title=title, description=description,article=article,interpretation=interpretation, enable_queue=enable_queue)
 iface.launch()
 
 #(num_top_classes=3)q