Aspose.OCR Version Benchmark
This benchmark compares the performance and accuracy of different versions of Aspose.OCR for .NET across various build configurations and runtime environments. The results demonstrate the evolution of OCR speed and accuracy improvements over multiple releases.
Dataset
- The benchmark was conducted on a dataset containing 1031 images.
- Each image was processed using various Aspose.OCR versions and configurations.
- Performance metrics include processing time and Levenshtein error percentage (Accuracy Error, %) as a measure of recognition accuracy.
Benchmark Results
Version | Build Target | Runtime Environment | Processing Time (ms) | Accuracy Error (%) | Performance Change |
|---|---|---|---|---|---|
v 23.12.0 | netstandard2.0 | NET 5.0 | 11468,85 | 20,67 | - |
v 24.11.0 | netstandard2.0 | NET 5.0 | 8508,62 | 5,71 | -35% |
v 25.7.0 | netstandard2.0 | NET 5.0 | 2545,82 | 5,29 | -234% |
v 25.11.0 | netstandard2.0 | NET 5.0 | 2518,27 | 5,30 | -1,09% |
v 25.11.0 | netstandard2.0 | NET 10.0 | 2360,47 | 5,30 | -6,69% |
- | NET 10.0 | NET 10.0 | 2348,14 | 5,30 | -0,52% |
- | NET 10.0 | NET 5.0 | 2557,25 | 5,30 | 8,18% |
- | NET 10.0 | Framework 4.7.2 | 2621,36 | 5,30 | 2,45% |
Amount of images: 1031
Table columns description:
- Version: Aspose.OCR version number being tested.
- Build Target: The .NET framework version used to build the library (netstandard2.0 or NET 10.0).
- Runtime Environment: The .NET runtime version on which the library was executed (NET 5.0, NET 10.0, or Framework 4.7.2).
- Processing Time (ms): Average processing time per image in milliseconds.
- Accuracy Error (%): Levenshtein error percentage - a measure of recognition accuracy (lower is better).
- Performance Change: Percentage change in processing time compared to the previous entry in the table.
Key Findings
- Performance Improvement: From v 23.12.0 to v 25.7.0, processing time decreased by approximately 78% (from 11468,85 ms to 2545,82 ms), representing a significant performance enhancement.
- Accuracy Improvement: The Levenshtein error rate improved dramatically from 20,67% in v 23.12.0 to 5,29% in v 25.7.0 and stabilized around 5,30% in later versions.
- Runtime Environment Impact: Running on NET 10.0 generally provides better performance compared to NET 5.0, with the best performance achieved using v 25.12.1 built for NET 10.0 and run on NET 10.0 (2348,14 ms).
- Build Target Impact: Versions built specifically for NET 10.0 show optimal performance when running on NET 10.0 runtime.
- Stability: Version 25.x series maintains consistent accuracy (5,30% Levenshtein error) while continuing to optimize performance.
Easy to use Aspose.OCR API
Aspose OCR API simplifies text extraction by handling image analysis and processing automatically, letting users focus on their core tasks—just provide an image and and leave the rest to the API.
Convert image to text
/** Initialize recognition API */
AsposeOcr api = new AsposeOcr();
/** Provide the image for recognition */
OcrInput source = new OcrInput(InputType.SingleImage);
source.Add("image.png");
/** Recognize and output results */
List<RecognitionResult> results = api.Recognize(source);
Console.WriteLine(results[0].RecognitionText);