Submissions are evaluated based on a combined measure of **classification** and **localization** accuracy. Your model must first determine whether pneumonia is present in each image, and for positive cases, accurately localize the affected regions with bounding boxes.

---

## Evaluation Metrics

**Classification Metrics:**

- F1 Score  
- Area Under the ROC Curve (AUC)

**Localization Metrics** (evaluated only for pneumonia-positive images):

- Precision  
- Recall  
- Intersection over Union (IoU) at a threshold of 0.5  
- mean Average Precision (mAP) 

The final score is computed as a weighted combination of classification and localization metrics.

---

## Submission File Format

Your submission must be a CSV file with the following columns:
- `patientId`: Filename of the test image (e.g. `bb393f7e-6ee6-473b-86e6-2e2dde4c8970`) 
- `predictionString`: Model predictions for the image

### `predictionString` format:

- If **no pneumonia** is predicted, fill this field in with **0.0 0.0 0.0 0.0 0.0**.
- If pneumonia **is predicted**, provide a space-separated string in the format: confidence x y width height


where:  
- `confidence` is the model’s confidence score (e.g., `0.95`)  
- `x`, `y` are the top-left coordinates of the bounding box   
- `width`, `height` are the dimensions of the bounding box 

If a patient has multiple bounding boxes, the space separated string will be confidence x y width height confidence x y width height

---

### Example Submission

```csv
patientId,predictionString
299d6232-d393-41bb-a531-9ef476496d4e, 
8ad2b94d-e34c-4fd3-b271-0de57b1e2e03, 0.7981 543.0 133.0 241.4 0.89 0.6258 567.4 15..
- Please download sample submission file under the Data section for more details