rlx-ocr 0.2.0

OCR engine for RLX — text detection + recognition
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152

// RLX — versatile ML compiler + runtime.
// Copyright (C) 2026 Eugene Hauptmann, Nataliya Kosmyna.
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, version 3.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with this program. If not, see <https://www.gnu.org/licenses/>.

//! High-level OCR runner with image loading.

use crate::capabilities::validate_device;
use crate::config::OcrConfig;
use crate::engine::{OcrEngine, OcrEngineParams};
use crate::text::TextLine;
use crate::weights::resolve_model_dir;
use anyhow::{Context, Result, anyhow};
use rlx_runtime::Device;
use rten_imageproc::RotatedRect;
use std::path::{Path, PathBuf};

/// Structured OCR output.
#[derive(Debug, Clone)]
pub struct OcrOutput {
    pub text: String,
    pub lines: Vec<Option<TextLine>>,
    pub words: Vec<RotatedRect>,
}

/// Builder for [`OcrRunner`] (mirrors whisper / dinov2 runners).
#[derive(Debug, Clone, Default)]
pub struct OcrRunnerBuilder {
    model_dir: Option<PathBuf>,
    detection_model: Option<PathBuf>,
    recognition_model: Option<PathBuf>,
    device: Option<Device>,
    alphabet: Option<String>,
}

impl OcrRunnerBuilder {
    pub fn model_dir<P: Into<PathBuf>>(mut self, dir: P) -> Self {
        self.model_dir = Some(dir.into());
        self
    }

    pub fn detection_model<P: Into<PathBuf>>(mut self, p: P) -> Self {
        self.detection_model = Some(p.into());
        self
    }

    pub fn recognition_model<P: Into<PathBuf>>(mut self, p: P) -> Self {
        self.recognition_model = Some(p.into());
        self
    }

    pub fn device(mut self, d: Device) -> Self {
        self.device = Some(d);
        self
    }

    pub fn alphabet(mut self, alphabet: impl Into<String>) -> Self {
        self.alphabet = Some(alphabet.into());
        self
    }

    pub fn build(self) -> Result<OcrRunner> {
        let (detection, recognition) = match (
            self.detection_model,
            self.recognition_model,
            self.model_dir,
        ) {
            (Some(d), Some(r), _) => (d, r),
            (_, _, Some(dir)) => resolve_model_dir(&dir)?,
            _ => {
                return Err(anyhow!(
                    "provide model_dir(...) or both detection_model(...) and recognition_model(...)"
                ));
            }
        };
        let device = self.device.unwrap_or(Device::Cpu);
        validate_device(device)?;

        let engine = OcrEngine::new(OcrEngineParams {
            detection_model: Some(detection),
            recognition_model: Some(recognition),
            alphabet: self.alphabet,
            device,
            ..Default::default()
        })?;

        Ok(OcrRunner { engine, device })
    }
}

/// OCR session wrapping a fully loaded [`OcrEngine`].
pub struct OcrRunner {
    engine: OcrEngine,
    device: Device,
}

impl OcrRunner {
    pub fn builder() -> OcrRunnerBuilder {
        OcrRunnerBuilder::default()
    }

    pub fn engine(&self) -> &OcrEngine {
        &self.engine
    }

    pub fn device(&self) -> Device {
        self.device
    }

    pub fn config(&self) -> OcrConfig {
        self.engine.config()
    }

    /// Run OCR on an image file (JPEG/PNG via `image` crate).
    pub fn predict_path(&self, path: &Path) -> Result<OcrOutput> {
        let img = image::open(path)
            .with_context(|| format!("open image {path:?}"))?
            .into_rgb8();
        let (w, h) = img.dimensions();
        self.predict_rgb(img.as_raw(), w, h)
    }

    /// Run OCR on RGB8 bytes.
    pub fn predict_rgb(&self, rgb: &[u8], width: u32, height: u32) -> Result<OcrOutput> {
        let source = crate::ImageSource::from_bytes(rgb, (width, height))?;
        let input = self.engine.prepare_input(source)?;
        let words = self.engine.detect_words(&input)?;
        let line_rects = self.engine.find_text_lines(&input, &words);
        let lines = self.engine.recognize_text(&input, &line_rects)?;
        let text = lines
            .iter()
            .filter_map(|l| l.as_ref().map(TextLine::text))
            .collect::<Vec<_>>()
            .join("\n");
        Ok(OcrOutput { text, lines, words })
    }

    /// Convenience: text only.
    pub fn predict_text(&self, path: &Path) -> Result<String> {
        Ok(self.predict_path(path)?.text)
    }
}