2 images sourced from the IAM Handwriting Database and the SRIOE dataset. These files serve as functional test fixtures for validating TrOCR vision-encoder-decoder models within the HuggingFace Transformers library.
Use Cases
- Validate the integration of TrOCR models using the test_modeling_vision_encoder_decoder_model.py script
- Benchmark OCR inference performance on standardized IAM and SRIOE image samples
- Debug vision-encoder-decoder architectures by comparing model outputs against the expected results defined in HuggingFace test fixtures
Strengths
- Contains 2 distinct images representing handwriting and structured document OCR tasks
- Includes one sample from the IAM Handwriting Database for cursive text recognition testing
- Includes one sample from the SRIOE dataset for receipt or structured text extraction testing