README.md

metadata

title: Lifespan Predictor
emoji: 📚
colorFrom: red
colorTo: pink
sdk: gradio
sdk_version: 5.44.1
app_file: app.py
pinned: false
license: mit
short_description: remaining lifespan predictor

🧑‍⚕️ Lifespan Prediction Demo

This is a Gradio app that predicts the remaining lifespan (in years) from a face image using a fine-tuned DINOv2 model.
The model was trained on the Remaining Lifespan Prediction Faces dataset.

---

🚀 How it works

• Upload a face photo or use your webcam.
• The model extracts features with a DINOv2 ViT-L backbone.
• A regression head outputs the predicted remaining lifespan (years).

---

🖥️ Usage

1. Click “Use Webcam” or “Upload Image”.
2. Wait a few seconds for processing (faster on GPU).
3. The result will be shown in the output box:
   Estimated remaining lifespan: XX.X years

---

⚙️ Technical Details

• Backbone: dinov2_vitl14_reg (frozen during training)
• Hidden dimension: 128
• Loss: Heteroscedastic Negative Log Likelihood (NLL)
• Best checkpoint saved during training → uploaded to Model Hub.

---

📦 Files in this Space

• app.py → Gradio app definition
• requirements.txt → dependencies
• README.md → this file

---

⚠️ Disclaimer

This demo is for research and educational purposes only.
Predictions are not medical advice and should not be used in real-world decision-making.

---