README.md

---
license: apache-2.0
tags:
- image-classification
- pytorch
- computer-vision
- ocr
- crossed-out-text
library_name: pytorch
pipeline_tag: image-classification
---

# crossed-out-text-classifier

## Model Description

This is a ResNet18-based binary classifier trained to detect crossed out text in OCR images. The model classifies images into two categories:
- `no`: Text is not crossed out
- `yes`: Text is crossed out

## Model Details

- **Architecture**: ResNet18 with modified classification head
- **Parameters**: 11,187,158
- **Input Size**: 224x224 RGB images
- **Classes**: ['no', 'yes']
- **Validation Accuracy**: 0.9688
- **Training Framework**: PyTorch

## Usage

### Using the model directly

```python
import torch
from PIL import Image
import torchvision.transforms as transforms

# Load model
model = torch.load('pytorch_model.bin', map_location='cpu')
model.eval()

# Prepare image
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

image = Image.open('your_image.png').convert('RGB')
input_tensor = transform(image).unsqueeze(0)

# Make prediction
with torch.no_grad():
    outputs = model(input_tensor)
    probabilities = torch.nn.functional.softmax(outputs, dim=1)
    predicted_class = torch.argmax(probabilities, dim=1).item()
    confidence = torch.max(probabilities, dim=1)[0].item()

class_names = ['no', 'yes']
print(f"Prediction: {class_names[predicted_class]} (confidence: {confidence:.4f})")
```

### Using the inference module

```python
from src.inference import CrossedOutPredictor

# Initialize predictor
predictor = CrossedOutPredictor()
predictor.load_model('pytorch_model.bin')

# Make prediction
prediction, confidence = predictor.predict_image('your_image.png')
print(f"Prediction: {prediction} (confidence: {confidence:.4f})")
```

## Training Data

The model was trained on a dataset of OCR images with crossed out and non-crossed out text. The training used:
- Data augmentation including rotation, scaling, shearing, and color jittering
- Transfer learning from ImageNet pretrained ResNet18
- Two-phase training: frozen backbone followed by full fine-tuning

## Limitations

- The model is specifically designed for OCR images and may not generalize well to other image types
- Performance may vary with different text fonts, sizes, or crossing-out patterns
- Trained on specific image resolution (224x224) and normalization

## Intended Use

This model is intended for:
- OCR post-processing pipelines
- Document analysis systems
- Text validation workflows

## License

This model is released under the Apache 2.0 license.

## Citation

If you use this model, please cite:

```bibtex
@misc{Sleeeepy_crossed_out_text_classifier,
  title={Crossed Out Text Classifier},
  author={Your Name},
  year={2025},
  howpublished={\url{https://huggingface.co/Sleeeepy/crossed-out-text-classifier}}
}
```