completed model

EffNetModel.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55228d3b8996e17611d32e845e3fdbae61c00ee8d7a4ac008a413a1da7734329
+size 31254650

app.py

@@ -0,0 +1,32 @@
+### Imports for Modules ### 
+import gradio as gr
+import os
+import torch
+from typing import Tuple, Dict
+from timeit import default_timer as timer
+
+### Functional Imports
+from predictor import predictionMaker
+
+exampleList = [["examples/" + example] for example in os.listdir("examples")]
+
+title = "Detecting Retinal Diseases for Early Prevention"
+description = "An EfficientNetB2 feature extractor computer vision model to classify OCT images into Brain Tumor types: CNV, DME, Drusen and Normal"
+article = "Created by [Eternal Bliassard](https://github.com/EternalBlissard)."
+
+# Create the Gradio demo
+demo = gr.Interface(fn=predictionMaker, 
+                    inputs=[gr.Image(type="pil")], 
+                    outputs=[gr.Label(num_top_classes=2, label="Predictions"),
+                             gr.Number(label="Prediction time (s)")],
+                    examples=exampleList, 
+                    title=title,
+                    description=description,
+                    article=article)
+
+# Launch the demo!
+demo.launch() 
+
+
+
+

helper.py

@@ -0,0 +1,36 @@
+
+import matplotlib.pyplot as plt
+import numpy as np
+import os
+import torch
+import random
+import zipfile
+from pathlib import Path
+import requests
+
+def setAllSeeds(seed):
+  os.environ['MY_GLOBAL_SEED'] = str(seed)
+  random.seed(seed)
+  np.random.seed(seed)
+  torch.manual_seed(seed)
+  torch.cuda.manual_seed_all(seed)
+  
+def dataDownloader(src,dest):
+  downloadPath = Path("downloadedData/")/dest
+
+  if(downloadPath.is_dir()):
+    print(f"{downloadPath} directory already exists, skipping downloading procedure")
+  else:
+    print(f"{downloadPath} directory doesn't already exists, starting downloading procedure")
+    downloadPath.mkdir(parents=True,exist_ok=True)
+    target = Path(src).name
+    with open(Path("downloadedData/")/target,"wb") as f:
+      requested = requests.get(src)
+      print(f"Downloading {target} from {src}")
+      f.write(requested.content)
+    
+    with zipfile.ZipFile(Path("downloadedData/")/target,"r") as zipRef:
+      print(f"Unzipping the data")
+      zipRef.extractall(downloadPath)
+      os.remove(Path("downloadedData/")/target)
+  return downloadPath

model.py

@@ -0,0 +1,18 @@
+
+import torch
+import torchvision
+from torch import nn
+from helper import setAllSeeds
+
+def getEffNetModel(seed,numClasses):
+  setAllSeeds(seed)
+  effNetWeights = torchvision.models.EfficientNet_B2_Weights.DEFAULT
+  effNetTransforms = effNetWeights.transforms()
+  effNet = torchvision.models.efficientnet_b2(weights=effNetWeights)
+  for param in effNet.parameters():
+    param.requires_grad = False
+  effNet.classifier = nn.Sequential(
+    nn.Dropout(p=0.3,inplace=True),
+    nn.Linear(1408,numClasses,bias=True)
+  )
+  return effNet,effNetTransforms

predictor.py

@@ -0,0 +1,24 @@
+### Imports for Modules ### 
+import gradio as gr
+import os
+import torch
+from typing import Tuple, Dict
+from timeit import default_timer as timer
+
+### Functional Imports
+from model import getEffNetModel
+
+classNames = ['CNV', 'DME', 'Drusen', 'Normal']
+effNetModel, effNetTransforms = getEffNetModel(42,len(classNames))
+effNetModel.load_state_dict(torch.load(f="EffNetModel.pt",map_location=torch.device("cpu")))
+
+def predictionMaker(img):
+  startTime = timer()
+  img = effNetTransforms(img).unsqueeze(0)
+  effNetModel.eval()
+  with torch.inference_mode():
+    predProbs = torch.softmax(effNetModel(img),dim=1)
+  predDict = {classNames[i]: float(predProbs[0][i]) for i in range(len(classNames))}
+  endTime = timer()
+  predTime = round(endTime-startTime,4)
+  return predDict,predTime

requirements.txt

@@ -0,0 +1,3 @@
+torch==2.2.0
+torchvision==0.17.0
+gradio==4.20.0