Add model card

README.md

@@ -1,26 +1,25 @@
-# ResNet50-APTOS-DR (ONNX)
+# ResNet50-APTOS-DR (ONNX) - Single Clean File
 
-**5-class Diabetic Retinopathy classifier** converted to ONNX for edge devices.
+**5-class Diabetic Retinopathy classifier** ready for Raspberry Pi 5 and edge devices.
 
 **Original model**: sakshamkr1/ResNet50-APTOS-DR  
-**Format**: ONNX (exported from PyTorch)  
-**Input**: RGB fundus image 224×224  
+**Format**: ONNX (single file - no external .data)  
+**Input shape**: (batch, 3, 224, 224) RGB fundus image  
 **Output**: 5 classes (APTOS 2019)
 
 ### Classes
-- **0**: No DR  
-- **1**: Mild DR  
-- **2**: Moderate DR  
-- **3**: Severe DR  
-- **4**: Proliferative DR  
+- 0: No DR  
+- 1: Mild DR  
+- 2: Moderate DR  
+- 3: Severe DR  
+- 4: Proliferative DR  
 
+### Perfect 
+- Model size: ~105 MB (single file)  
+- RAM usage: ~150-220 MB  
+- Speed: ~0.8–1.5 seconds per image on CPU
 
-
-### How to use this model
-
-```bash
-pip install onnxruntime pillow torchvision
-```
+### Quick test code for Pi 5
 
 ```python
 import onnxruntime as ort
@@ -28,7 +27,8 @@ import numpy as np
 from PIL import Image
 import torchvision.transforms as transforms
 
-session = ort.InferenceSession("iris.onnx", providers=["CPUExecutionProvider"])
+session = ort.InferenceSession("iris-vit.onnx", 
+                               providers=["CPUExecutionProvider"])
 
 transform = transforms.Compose([
     transforms.Resize((224, 224)),
@@ -36,19 +36,17 @@ transform = transforms.Compose([
     transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
 ])
 
-img = Image.open("your_photo.jpg").convert("RGB")
+img = Image.open("DR1.jpg").convert("RGB")
 input_tensor = transform(img).unsqueeze(0).numpy().astype(np.float32)
 
 output = session.run(None, {"input": input_tensor})[0][0]
 probs = np.exp(output) / np.sum(np.exp(output))
-pred = np.argmax(probs)
+pred_idx = np.argmax(probs)
 
-class_names = ["No DR", "Mild DR", "Moderate DR", "Severe DR", "Proliferative DR"]
-print(f"Prediction: {class_names[pred]} ({probs[pred]*100:.1f}%)")
+classes = ["No DR", "Mild DR", "Moderate DR", "Severe DR", "Proliferative DR"]
+print(f"✅ Predicted: {classes[pred_idx]} ({probs[pred_idx]*100:.1f}%)")
 ```
 
-### License
-MIT  
-**Feel free to use in any project (research / commercial)**
+**License**: MIT  
+Made for low-resource diabetic retinopathy screening ❤️
 
-Made with ❤️ for low-resource DR screening.