GitLab

- Add support for Apple M1 chips (mps device)
- Refactor image processing for better performance and compatibility
- Update model loading and inference for various devices
- Adjust batch processing and memory management

- Add UnimerMBartConfig and UnimerSwinConfig classes
- Implement UnimerSwinImageProcessor for image preprocessing- Create necessary __init__.py files for module structure

- Move title level determination to the beginning of the Title block processing
- Add condition to include text_level only if it's not 0
- Adjust title level to 0 instead of 1 when it's less than 1

- Remove unnecessary __getitem__ method
- Simplify image cropping in detect_math_formula_region
- Improve code readability and efficiency

- Set NPUDTCompile to false for better performance on NPU
- Adjust batch ratio

- Include BlockType.Discarded in the list of compatible block types for ContentType.Text and ContentType.InlineEquation
- This change improves the OCR dictionary merging process by handling discarded blocks more effectively

- Remove unused @ImportPIL decorator from load_images_from_pdf function
- Update image shape handling in YOLOv11.py for better compatibility

These changes improve code readability and performance without altering the original functionality.

- Replace PIL with cv2 for image processing
- Fix issues with image cropping and resizing
- Add boundary checks and error handling
- Optimize code for better performance and readability

- Remove PIL usage across multiple files
- Convert image processing functions to use NumPy arrays
- Update crop_img function to work with NumPy arrays
- Modify image loading and resizing to use NumPy and OpenCV
- Clean up unused imports and comments related to PIL

- Optimize paragraph splitting algorithm for better text block separation
- Update fast-langdetect dependency to ensure compatibility

- Increase batch ratio to 8 for GPU memory >=16GB
- Improve inference performance on systems with higher GPU memory

- Update OCR dictionary merge logic to include text blocks when processing interline equations
- This change improves the handling of equations that may be embedded within text content

- Simplify batch ratio logic for GPU memory >= 16GB
- Remove unnecessary conditions for 20GB and 40GB memory

- Simplify batch ratio logic for GPU memory >= 16GB
- Remove unnecessary conditions for 20GB and 40GB memory

- Sort detected images by area before processing to enhance MFR accuracy
- Implement stable sorting to maintain original order of images with equal

- Comment out @measure_time decorator for txt_spans_extract_v2 and sort_lines_by_model functions
- Remove logger.info for page_process_time
- Comment out PerformanceStats.print_stats call

- Add performance_stats module to measure and print execution time statistics
- Implement measure_time decorator to track execution time of key functions
- Remove multi-threading in pdf parsing for better resource management
- Optimize pdf parsing logic for improved performance

- Add ThreadPoolExecutor to process PDF pages in parallel
- Create separate function for page processing to improve readability and maintainability
- Include error handling for individual page processing tasks
- Log total page processing time for performance monitoring

- Update condition to only convert full-width letters and numbers
- Remove separate case for full-width space

- Replace complex device selection logic with a single line using torch.device
- Remove redundant checks and imports for better readability and maintainability

- Remove supports_bfloat16 variable and related checks
- Remove model.bfloat16() call for LayoutLMv3ForTokenClassification
- Simplify device selection logic

- Implement full_to_half function to convert full-width characters to half-width
- Apply conversion to span content before merging paragraphs
- Improve text processing for better readability and consistency

- Move batch model initialization outside the loop
- Collect page dimensions before analyzing- Update page info dictionary structure
- Add null dimensions for non-analyzed pages

- Introduce span_block_type_compatible function to check compatibility between span and block types
- Update fill_spans_in_blocks function to use the new compatibility check
- Improve accuracy of span allocation to blocks based on content type

- Update the logic for determining `end_page_id` to handle negative values
- This change ensures proper behavior when `end_page_id` is set to -1 or other negative values

- Add license ID information to the log for better traceability
- Improve logging format to include both license ID and expiration date

- Add ImportError handling to silence known import-related exceptions
- Improve generic exception handling to log error messages- Maintain existing specific exception handlers for license-related issues