initial commit

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+pretrained_effnetb2_feature_extractor_20_percent.pth filter=lfs diff=lfs merge=lfs -text
+.pth filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,75 @@
+import gradio as gr
+import os
+import torch
+from model import create_effnetb2_model
+from timeit import default_timer as timer
+from typing import Tuple, Dict
+
+
+# Define the class names
+class_names = ["pizza", "steak", "sushi"]
+
+# Create the pretrained EffNetB2 model
+effnetb2, effnetb2_transforms = create_effnetb2_model(
+    num_classes=3,
+)
+
+# Load to the CPU the EffNetB2 model's saved weights
+effnetb2.load_state_dict(
+    torch.load(
+        f="pretrained_effnetb2_feature_extractor_20_percent.pth",
+        map_location=torch.device("cpu"),
+    )
+)
+
+
+def predict(img) -> Tuple[Dict, float]:
+    """This function transforms and performs a prediction on an image, and returns
+    the prediction and time taken.
+
+    Returns: the prediction dictionary and prediction time.
+    """
+    # Begin the prediction's timer
+    start_time = timer()
+
+    # Transform the target image with the pretrained EffNetB2 model's transforms, and add a batch dimension
+    img = effnetb2_transforms(img).unsqueeze(0)
+
+    # Set the model to evaluation model
+    effnetb2.eval()
+
+    # Activate the inference mode
+    with torch.inference_mode():
+        # Pass the transformed image through the model, and transform the prediction logits (the model's outputs) into prediction probabilities
+        pred_probs = torch.softmax(effnetb2(img), dim=1)
+
+    # Create a prediction label and prediction probability dictionary for each prediction class
+    pred_labels_and_probs = {class_names[i]: float(pred_probs[0][i]) for i in range(len(class_names))}
+
+    # Calculating the prediction's time
+    pred_time = round(timer() - start_time, 5)
+
+    return pred_labels_and_probs, pred_time
+
+
+# Define the title
+title = "DishVision - Multi Class Food Image Classifier"
+# Define the description
+description = "An EfficientNetB2-based transfer learning model for feature extraction in computer vision, designed to classify images into three distinct food categories."
+# Define the article
+article = "Computer Vision Project"
+
+# Create the "examples" list from "examples/" directory
+example_list = [["examples/" + example] for example in os.listdir("examples")]
+
+# Create the Gradio deom interface
+demo = gr.Interface(fn=predict,
+                    inputs=gr.Image(type="pil"),
+                    outputs=[gr.Label(num_top_classes=3, label="Predictions"),
+                             gr.Number(label="Prediction time (s)")],
+                    examples=example_list,
+                    title=title,
+                    description=description,
+                    article=article)
+
+demo.launch(share=True)

model.py

@@ -0,0 +1,39 @@
+import torch
+import torchvision
+from torch import nn
+
+
+def create_effnetb2_model(num_classes:int=3,
+                          seed:int=42):
+    """This function creates an EfficientNetB2 feature extractor model and transforms.
+
+    Args:
+        num_classes (int, optional): number of classes in the classifier head.
+            Defaults to 3.
+        seed (int, optional): random seed value. Defaults to 42.
+
+    Returns:
+        model (torch.nn.Module): EffNetB2 feature extractor model.
+        transforms (torchvision.transforms): EffNetB2 image transforms.
+    """
+    # Get the pretrained 'efficientnet_b2' model's weights
+    weights = torchvision.models.EfficientNet_B2_Weights.DEFAULT
+    # Get the pretrained 'efficientnet_b2' model's transforms
+    transforms = weights.transforms()
+    # Set up the pretrained 'efficientnet_b2' model
+    model = torchvision.models.efficientnet_b2(weights=weights)
+
+    # Freeze the base layers in the 'efficientnet_b2' model (for Feature Extraction)
+    for param in model.parameters():
+        param.requires_grad = False
+
+    # Set the torch manual seed
+    torch.manual_seed(seed)
+
+    # Update the classifier head to suit to our problem
+    model.classifier = nn.Sequential(
+        nn.Dropout(p=0.3, inplace=True),
+        nn.Linear(in_features=1408, out_features=num_classes),
+    )
+
+    return model, transforms

pretrained_effnetb2_feature_extractor_20_percent.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a14a8d707956ac7a4ced4ee884df9c92850795ea18ac3b5d41c3c34dbc614253
+size 31298682

requirements.txt

@@ -0,0 +1,4 @@
+torch==2.0.1
+torchvision==0.15.2
+gradio==4.38.1
+numpy<2.0.0