| Property | Value |
|---|---|
| Parameter Count | 609M |
| Model Type | Vision-Encoder-Decoder |
| Tensor Type | F32 |
| Base Model | microsoft/trocr-large-printed |
What is trocr-large-printed-cmc7_tesseract_MICR_ocr?
This is a specialized vision-encoder-decoder model fine-tuned from Microsoft's TrOCR large printed model. It's specifically designed for optical character recognition (OCR) tasks, with a focus on CMC7 and MICR text formats commonly used in banking and financial documents.
Implementation Details
The model was trained using the Adam optimizer with carefully tuned hyperparameters (β1=0.9, β2=0.999, ε=1e-08) and implements a linear learning rate scheduler. Training was conducted over 5 epochs with a learning rate of 5e-05 and batch sizes of 16 for both training and evaluation.
- Built on Transformers 4.39.3 framework
- Utilizes PyTorch 2.1.2
- Incorporates Datasets 2.18.0
- Uses Tokenizers 0.15.2
Core Capabilities
- Specialized text recognition for printed documents
- Optimized for CMC7 and MICR character formats
- Supports TensorBoard integration for monitoring
- Compatible with Inference Endpoints
Frequently Asked Questions
Q: What makes this model unique?
This model specializes in recognizing printed text, particularly focusing on specialized formats like CMC7 and MICR, making it particularly valuable for financial document processing and banking applications.
Q: What are the recommended use cases?
The model is best suited for OCR tasks involving printed documents, especially those containing standardized financial text formats. It's particularly useful for processing bank checks, financial statements, and other documents using CMC7 or MICR encoding.