Scope and Limitations

This page lists what Aspose.OCR FOSS for .NET supports and does not support today, so you can confirm compatibility before integrating it. The public API is intentionally compact and may evolve as recognition capabilities are extended.

Language and Script

Recognition supports Latin-script text, including the diacritical characters present in the embedded recognition model. There is no runtime language-selection API — the recognized alphabet is fixed by the embedded model.

Input Types

Inputs are limited to InputType.SingleImage (raster images supported by System.Drawing.Common) and InputType.PDF. Any other input type throws NotSupportedException. PDF pages are always rendered at 150 DPI; this resolution is not configurable.

Detection Output

The detector returns axis-aligned text-line rectangles only. Polygonal regions and layout labels (headings, tables, columns) are not part of the public output.

Not Implemented

These capabilities are not implemented in Aspose.OCR FOSS for .NET:

Searchable PDF generation or text-export helpers
Confidence scores for recognized text
Font detection
Spelling correction
Table structure detection
Barcode detection
Automatic page rotation, deskewing, or color inversion
User-configurable preprocessing

Platform and Execution

Image processing is Windows-only, because the pipeline depends on System.Drawing.Common for bitmap loading, cropping, resizing, and pixel access.
Inference runs on the CPU only — GPU execution is not configured.
Input files and PDF pages are processed sequentially; there is no input-level multithreading API, although Microsoft.ML.OnnxRuntime parallelizes work inside each inference call.

Assembly Size

The compiled assembly is large because all three ONNX models (detection, refinement, and recognition) are embedded as resources in every target-framework build.

Scope and Limitations