README.md

---
license: mit
language:
- en
metrics:
- precision
- recall
- f1
base_model:
- openmmlab/upernet-swin-base
tags:
- image-segmentation
- text-segmentation
- text-detection
- scene-text-detection
---

# Model Card for indic-text-detection

This model preicts the binary map for a given scene image. The binary map shows the location of text regions.

## Model Details

This model extends `openmmlab/upernet-swin-base` and performs binarization which generates the segmentation map.


![Proposed_Model.jpg](https://cdn-uploads.huggingface.co/production/uploads/6868f8219c4cd7445653ada1/d0hK3K7xPY3MfIr_0ynB0.jpeg)

## Example

![Detection_Anamoly_2.jpg](https://cdn-uploads.huggingface.co/production/uploads/6868f8219c4cd7445653ada1/XgtugD9mHrWnggbGRJjbv.jpeg)

## Evaluation Metrics

| **Metric**  | **Tiny** | **Base** |
|-------------|----------|----------|
| Precision   | 0.8352   | 0.8628   |
| Recall      | 0.8411   | 0.8521   |
| F-Score     | 0.8381   | 0.8574   |