Tokenized-OCR

Tokenized-OCR is an advanced OCR-based text extraction tool optimized for generating structured, tokenized outputs. Built upon a powerful vision-language architecture with enhanced OCR and multilingual support, Tokenized-OCR accurately extracts text from images and returns it as a comma-separated sequence.

Key Enhancements:

• Advanced OCR Engine: Fine-tuned on extensive datasets, Tokenized-OCR ensures precise text recognition and tokenization.
• Optimized for Tokenized Output: Produces structured comma-separated text, making it ideal for downstream NLP tasks, automation pipelines, and database integrations.
• Enhanced Multilingual OCR: Supports text extraction in multiple languages, including English, Chinese, Japanese, Korean, Arabic, and more.
• Multimodal Processing: Seamlessly processes both image and text inputs, providing structured tokenized outputs.
• Secure and Optimized Model Weights: Employs safetensors for efficient and secure model loading.

Demo Inference

Instruction : "Extract and return the tokenized OCR text from the image, ensuring separated by commas."

How to Use

from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info

# Load the Tokenized-OCR model with optimized parameters
model = Qwen2VLForConditionalGeneration.from_pretrained(
    "prithivMLmods/Tokenized-OCR", torch_dtype="auto", device_map="auto"
)

# Recommended acceleration for performance optimization:
# model = Qwen2VLForConditionalGeneration.from_pretrained(
#     "prithivMLmods/Tokenized-OCR",
#     torch_dtype=torch.bfloat16,
#     attn_implementation="flash_attention_2",
#     device_map="auto",
# )

# Load the default processor for Tokenized-OCR
processor = AutoProcessor.from_pretrained("prithivMLmods/Tokenized-OCR")

# Define the input messages with both an image and a text prompt
messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "image",
                "image": "https://flux-generated.com/sample_image.jpeg",
            },
            {"type": "text", "text": "Extract and return the tokenized OCR text from the image, ensuring each word is accurately recognized and separated by commas."},
        ],
    }
]

# Prepare the input for inference
text = processor.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True
)
image_inputs, video_inputs = process_vision_info(messages)
inputs = processor(
    text=[text],
    images=image_inputs,
    videos=video_inputs,
    padding=True,
    return_tensors="pt",
)
inputs = inputs.to("cuda")

# Generate the output
generated_ids = model.generate(**inputs, max_new_tokens=256)
generated_ids_trimmed = [
    out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
    generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text)

Key Features

1. High-Accuracy OCR Processing

   • Extracts and tokenizes text from images with exceptional precision.

2. Multilingual Text Recognition

   • Supports multiple languages, ensuring comprehensive OCR capabilities.

3. Comma-Separated Tokenized Output

   • Generates structured text for seamless NLP and data processing tasks.

4. Efficient Image & Text Processing

   • Handles both visual and textual inputs, ensuring accurate OCR-based extraction.

5. Optimized for Secure Deployment

   • Uses safetensors for enhanced security and model efficiency.

Tokenized-OCR revolutionizes text extraction from images, providing tokenized outputs that are easy to integrate into automated workflows, search engines, and language processing applications.