Uploading the main app files

DenseNet121d_22_From_Scratch_model0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a62ef98a230da5ece3c44c04fc24b955a362740c3fbbe2b0d09b37b719b06dc7
+size 28571079

app.py

@@ -0,0 +1,74 @@
+### 1. Imports and class names setup ###
+import gradio as gr
+import os
+import torch
+
+from model import create_DenseNet121_model
+from timeit import default_timer as timer
+from typing import Tuple, Dict
+
+# Setup class names
+class_names = ['infected', 'notinfected']
+
+### 2. Model and transforms preparation ###
+
+# Create an instance of trained DenseNet121 model
+Dense121, transform = create_DenseNet121_model()
+
+### 3. Predict function ###
+
+# Create predict function
+def predict(img) -> Tuple[Dict, float]:
+  """
+  Transforms and performs a prediction on img then returns prediction and time taken
+  """
+  # Start the timer
+  start_time = timer()
+
+  # Transform the target image and add a batch dimension
+  img = transform(img).unsqueeze(0)
+  
+  # Put model into evaluation mode and turn on the inference mode
+  Dense121.eval()
+  with torch.inference_mode():
+    # Pass the transformed image through the model and turn the prediction logit intp prediction probability
+    pred_logit = Dense121(img).squeeze()
+    pred_prob = torch.sigmoid(pred_logit)
+    pred_label = torch.round(pred_prob)
+    pred_label = pred_label.type(torch.int64)
+    pred_class = class_names[pred_label.cpu()]
+    # pred_prob = float(pred_prob) # This line and next one are for formatting the pred_prob to print only 4 decimal places
+    # pred_prob = round(pred_prob, 4)
+
+  # Create a prediction label and prediction probability dictionary for each prediction class (this is the required format for Gradio's output parameter)
+  pred_label = pred_class 
+
+  # Calculate the prediction time
+  pred_time = round(timer() - start_time, 5)
+
+  # Return the prediction dictionary and prediction time
+  return pred_label, pred_time
+
+### 4. Gradio app ###
+
+# Create title and description strings
+title = "PCOS Detector in Ultrasound Images"
+description = "A DenseNet121 feature extractor computer vision model trained from scratch to classify ultrasound images of ovaries into PCOS infected or not infected."
+article= "Code implementation available at [GitHub](https://github.com/haidary99?tab=repositories)"
+
+# Create examples list from "examples/" directory
+example_list = [["examples/" + example] for example in os.listdir("examples")]
+
+# Create the Gradio demo
+demo = gr.Interface(fn=predict, # mapping function from input to output
+                    inputs=gr.Image(type="pil"),
+                    outputs=[gr.Label(label="Model Prediction"),
+                             gr.Number(label="Prediction time (s)")],
+                    # Create examples list from "examples/" directory
+                    examples=example_list,
+                    title=title,
+                    description=description,
+                    article=article)
+
+# Launch the demo
+demo.launch()

model.py

@@ -0,0 +1,19 @@
+#!pip install timm==0.6.13
+import torch
+import timm
+
+from torchvision import transforms
+
+def create_DenseNet121_model(): # Returns trained DenseNet121 model and its transforms
+  model_file = "DenseNet121d_22_From_Scratch_model0.pth"
+  model = torch.load(model_file, map_location=torch.device('cpu'))
+
+  transform = transforms.Compose([
+    transforms.Resize((224, 224)), # 1. Reshape all images to 224x224
+    transforms.ToTensor(), # Turn pixel values to between 0 & 1
+    transforms.Normalize(mean=[0.485, 0.456, 0.406], # 3. A mean of [0.485, 0.456, 0.406] (across each colour channel)
+                         std=[0.229, 0.224, 0.225]), # 4. A standard deviation of [0.229, 0.224, 0.225] (across each colour channel)
+    transforms.Grayscale() ##### change number of color channels from 3 to 1 (I added this)
+  ])
+
+  return model, transform

requirements.txt

@@ -0,0 +1,4 @@
+torch==1.12.0
+torchvision==0.13.0
+gradio==3.1.4
+timm==0.6.13