Api Reference

TextResult & Output Modes

Public reference for input sources, output shapes, bounding polygon coordinates, and text frames.

Every OCR detection execution processes one ImageSource and returns either structured bounding blocks or reconstructed spatial text.

📥 Input Image Sources

Binding	File / URI Source	In-Memory Buffer Source	Notes
Rust	`ImageSource::Path(PathBuf)`	`ImageSource::Bytes(Vec<u8>)`	In-memory decoding uses pure Rust `image` engine.
Python	`ImageSource.PATH(str)`	`ImageSource.BYTES(bytes)`	Native UniFFI bindings.
Node.js	`detectTextFile(path)`	`detectTextData(Buffer)`	Native N-API bindings.
Java	`detectTextFile(path)`	`detectTextData(byte[])`	JNA dynamic linking against `librusto`.
Golang	`DetectTextFile(path)`	`DetectTextData([]byte)`	Native Cgo bindings against `librusto`.
Flutter/Dart	`detectTextFile(path)`	`detectTextData(Uint8List)`	Pure Dart FFI & Flutter platform plugin.
.NET	`UriImageSource(string)`	`BytesImageSource(byte[])`	Accepts filesystem paths or `file:` URIs.
Swift	`.uri(String)`	`.bytes(Data)`	Accepts filesystem paths or file URL strings.
Kotlin	`ImageSource.Uri(String)`	`ImageSource.Bytes(ByteArray)`	Natively resolves `content://` URIs and `file://` paths.
React Native	`{ uri: string }`	`{ base64: string }` or `{ bytes: Uint8Array }`	Exactly one key required in object payload.

📐 Coordinate Space & Geometry

Coordinates are expressed in image pixel units relative to the top-left origin (0, 0):

Rendering diagram...

box_points: Exactly 4 quadrilateral polygon vertices: [Top-Left, Top-Right, Bottom-Right, Bottom-Left]. Follows true text rotation angle.
frame: Axis-aligned bounding rectangle: (left, top, width, height) computed from the polygon bounds.

📤 Output Modes Breakdown

Mode	Return Shape	Use Case
`lines` (Default)	Structured List (`TextResult[]`)	Standard document reading, paragraph parsing, and form field extraction line by line.
`words`	Structured List (`TextResult[]`)	Interactive word tapping, word-level highlighting, or granular bounding box overlays.
`spatial`	Plaintext String (`string`)	Preserves 2D columns, tabular alignment, receipt rows, and document indentations.

🗂️ Structured Data Model (`TextResult`)

[
  {
    "text": "Total Due",
    "score": 0.992,
    "box_points": [
      [40.0, 120.0],
      [140.0, 120.0],
      [140.0, 148.0],
      [40.0, 148.0]
    ],
    "frame": {
      "left": 40.0,
      "top": 120.0,
      "width": 100.0,
      "height": 28.0
    }
  },
  {
    "text": "$250.00",
    "score": 0.987,
    "box_points": [
      [280.0, 120.0],
      [360.0, 120.0],
      [360.0, 148.0],
      [280.0, 148.0]
    ],
    "frame": {
      "left": 280.0,
      "top": 120.0,
      "width": 80.0,
      "height": 28.0
    }
  }
]

Type Mapping Across Bindings

Concept	Rust	.NET / C#	Swift (iOS)	Kotlin (Android)	React Native
Collection	`Vec<TextResult>`	`IReadOnlyList<TextResult>`	`[TextResult]`	`List<TextResult>`	`TextResult[]`
Text String	`item.text`	`item.Text`	`item.text`	`item.text`	`item.text`
Confidence	`item.score`	`item.Score`	`item.score`	`item.score`	`item.score`
Polygon Quad	`item.box_points`	`item.BoxPoints`	`item.boxPoints`	`item.boxPoints`	`item.box_points`
Frame	`item.frame`	`item.Frame`	`item.frame`	`item.frame`	`item.frame`

💻 Result Handling Examples

Rust

match ocr.detect_text(&source, &options)? {
    DetectTextResult::Structured(items) => {
        for item in items {
            println!("{} ({:.1}%) at [{}, {}]", item.text, item.score * 100.0, item.frame.left, item.frame.top);
        }
    }
    DetectTextResult::Spatial(spatial_text) => {
        println!("{spatial_text}");
    }
}

.NET / C#

var result = ocr.DetectText(source, options);

if (result is StructuredDetectTextResult structured)
{
    foreach (var item in structured.Items)
    {
        Console.WriteLine($"{item.Text} ({item.Score:P0}) [{item.Frame.Left}, {item.Frame.Top}]");
    }
}
else if (result is SpatialDetectTextResult spatial)
{
    Console.WriteLine(spatial.Text);
}

iOS (Swift)

let result = try ocr.detectText(source, options: options)

switch result {
case .structured(let items):
    for item in items {
        print("\(item.text) (\(item.score)) [\(item.frame.left), \(item.frame.top)]")
    }
case .spatial(let text):
    print(text)
}

Android (Kotlin)

when (val result = ocr.detectText(source, options)) {
    is DetectTextResult.Structured -> {
        result.items.forEach { item ->
            println("${item.text} (${item.score}) [${item.frame.left}, ${item.frame.top}]")
        }
    }
    is DetectTextResult.Spatial -> {
        println(result.text)
    }
}

React Native

// Lines output -> TextResult[]
const lines = await detectText({ uri }, { output: 'lines' });
lines.forEach((item) => console.log(item.text, item.frame));

// Spatial output -> string
const spatialText = await detectText({ uri }, { output: 'spatial' });
console.log(spatialText);

Edit this pageorReport an issue

OcrRunOptions

Request-local output options, resize bounds, DBNet detector settings, and postprocessing reference.