---
language: en
license: mit
tags:
  - medical-imaging
  - pneumonia-detection
  - chest-xray
  - tensorflow
  - transfer-learning
  - vgg16
metrics:
  - accuracy
  - auc
  - f1
pipeline_tag: image-classification
---

# Pneumonia Detection Model (VGG16)

A VGG16 model fine-tuned for binary classification of chest X-ray images.

## Model Performance
| Metric | Value |
|--------|-------|
| AUC-ROC | 0.8679 |
| Accuracy | 0.7679 |
| F1-Score | 0.6156 |
| Recall | 0.8248 |
| Specificity | 0.7514 |
| Precision | 0.4911 |

## Training Details
- **Base Model:** VGG16 (ImageNet pre-trained)
- **Fine-tuning:** 2-phase (feature extraction + fine-tuning)
- **Dataset:** RSNA Pneumonia Detection Challenge (26,684 patients)
- **Input Size:** 224x224x3
- **Optimal Threshold:** 0.4961 (Youden's J)

## Usage
```python
import tensorflow as tf
from huggingface_hub import hf_hub_download

model_path = hf_hub_download(repo_id="vksfdc2025/pneumonia-detection-model", filename="best_pneumonia_model.keras")
model = tf.keras.models.load_model(model_path)
```

## Live Demo
Try the model: [Hugging Face Space](https://huggingface.co/spaces/vksfdc2025/pneumonia-detection-xray)

## Author
Vamsee Krishna Kondapalli | AIML Program Capstone Project