Upload 3 files

README.md

@@ -1,3 +1,66 @@
 ---
 license: apache-2.0
+library_name: pytorch
+pipeline_tag: image-classification
+tags:
+- medical-imaging
+- mri
+- alzheimer
+- convnext
+- deep-learning
 ---
+
+# Alzheimer MRI ConvNeXt Classifier
+
+This repository contains a **GPU-accelerated deep learning model** for classifying Alzheimer’s disease stages from brain MRI images using a **ConvNeXt-based architecture**.
+
+The model is designed for **research, educational use, and technical demonstrations**, and is deployed as a **Hugging Face Inference Endpoint** for fast GPU inference.
+
+---
+
+## 🧠 Model Overview
+
+- **Task:** MRI image classification  
+- **Modality:** Brain MRI (2D slices)  
+- **Architecture:** ConvNeXt  
+- **Framework:** PyTorch  
+- **Deployment:** Hugging Face GPU Inference Endpoint  
+
+The model predicts probabilities over predefined Alzheimer-related classes provided in `class_names.json`.
+
+---
+
+## 📦 Repository Structure
+
+├── inference.py # Hugging Face inference entrypoint
+├── requirements.txt # Minimal runtime dependencies
+├── models/
+│ ├── best_model.pth # Trained ConvNeXt weights
+│ └── class_names.json # Class index → label mapping
+└── README.md
+
+
+---
+
+## ⚡ Inference
+
+The model is exposed via a **Hugging Face Inference Endpoint** and accepts an image file as input.
+
+### Example API Call
+
+```python
+import requests
+
+API_URL = "https://<your-endpoint>.endpoints.huggingface.cloud"
+HEADERS = {
+    "Authorization": "Bearer YOUR_HF_TOKEN"
+}
+
+with open("sample_mri.png", "rb") as f:
+    response = requests.post(
+        API_URL,
+        headers=HEADERS,
+        files={"file": f}
+    )
+
+print(response.json())

inference.py

@@ -0,0 +1,41 @@
+import json
+import torch
+import timm
+from PIL import Image
+import torchvision.transforms as T
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+# Load class names
+with open("models/class_names.json") as f:
+    CLASS_NAMES = json.load(f)
+
+# Load model
+model = timm.create_model(
+    "convnext_base",
+    pretrained=False,
+    num_classes=len(CLASS_NAMES)
+)
+
+state = torch.load("models/best_model.pth", map_location=device)
+model.load_state_dict(state)
+model.to(device)
+model.eval()
+
+transform = T.Compose([
+    T.Resize((224, 224)),
+    T.ToTensor(),
+    T.Normalize(mean=[0.5], std=[0.5])
+])
+
+def predict(image: Image.Image):
+    x = transform(image).unsqueeze(0).to(device)
+
+    with torch.no_grad():
+        logits = model(x)
+        probs = torch.softmax(logits, dim=1)[0]
+
+    return {
+        CLASS_NAMES[str(i)]: float(probs[i])
+        for i in range(len(CLASS_NAMES))
+    }

requirements.txt

@@ -0,0 +1,4 @@
+torch
+torchvision
+timm
+Pillow