|
|
--- |
|
|
language: en |
|
|
tags: |
|
|
- handwriting-recognition |
|
|
- vision2seq |
|
|
- qwen |
|
|
- image-to-text |
|
|
- htr |
|
|
- tensorflow |
|
|
license: mit |
|
|
pipeline_tag: image-to-text |
|
|
library_name: transformers |
|
|
--- |
|
|
|
|
|
# 🖋️ Finetuned Full HTR Model (Qwen-based) |
|
|
|
|
|
This is a **Qwen Vision2Seq** model fine-tuned for **Handwritten Text Recognition (HTR)**. It reads handwritten text from images and generates clean, editable output using advanced transformer-based image-to-text techniques. |
|
|
|
|
|
## 🔍 Model Summary |
|
|
|
|
|
- **Model Architecture**: Qwen-Vision2Seq (Image encoder + Language decoder) |
|
|
- **Framework**: TensorFlow (via Hugging Face Transformers) |
|
|
- **Input**: Handwritten text image |
|
|
- **Output**: Recognized plain text |
|
|
|
|
|
## 🧠 How to Use (with Hugging Face Transformers) |
|
|
|
|
|
```python |
|
|
from transformers import AutoProcessor, AutoModelForVision2Seq |
|
|
from PIL import Image |
|
|
import torch |
|
|
|
|
|
# Load processor and model |
|
|
processor = AutoProcessor.from_pretrained("Emeritus-21/Finetuned-full-HTR-model", trust_remote_code=True) |
|
|
model = AutoModelForVision2Seq.from_pretrained("Emeritus-21/Finetuned-full-HTR-model", trust_remote_code=True) |
|
|
|
|
|
device = "cuda" if torch.cuda.is_available() else "cpu" |
|
|
model = model.to(device) |
|
|
|
|
|
# Load and process image |
|
|
image = Image.open("your_image.jpg").convert("RGB") |
|
|
inputs = processor(images=image, return_tensors="pt").to(device) |
|
|
|
|
|
# Generate prediction |
|
|
generated_ids = model.generate(**inputs) |
|
|
recognized_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0] |
|
|
|
|
|
print("📝 Recognized Text:", recognized_text) |
|
|
|
