Add RTL text direction support for 0.8.1

Cargo.toml

@@ -3,7 +3,7 @@ members = [".", "oar-ocr-derive", "oar-ocr-core", "oar-ocr-vl"]
 resolver = "2"
 
 [workspace.package]
-version = "0.8.0"
+version = "0.8.1"
 edition = "2024"
 rust-version = "1.95"
 license = "Apache-2.0"
@@ -12,8 +12,8 @@ repository = "https://github.com/greatv/oar-ocr"
 homepage = "https://github.com/greatv/oar-ocr"
 
 [workspace.dependencies]
-oar-ocr-core = { version = "0.8.0", path = "oar-ocr-core", default-features = false }
-oar-ocr-derive = { version = "0.8.0", path = "oar-ocr-derive", default-features = false }
+oar-ocr-core = { version = "0.8.1", path = "oar-ocr-core", default-features = false }
+oar-ocr-derive = { version = "0.8.1", path = "oar-ocr-derive", default-features = false }
 
 # Shared third-party dependencies (kept in sync via workspace inheritance).
 clap = { version = "4.5.42", features = ["derive"] }

oar-ocr-core/Cargo.toml

@@ -56,6 +56,7 @@ thiserror = "2.0"
 tokenizers.workspace = true
 tracing.workspace = true
 tracing-subscriber.workspace = true
+unicode-bidi = "0.3.18"
 ureq = { version = "3.0", default-features = false, features = ["rustls", "platform-verifier"], optional = true }
 wide = { version = "1.5", optional = true }
 

oar-ocr-core/src/domain/adapters/text_recognition_adapter.rs

@@ -8,7 +8,10 @@ use crate::core::traits::{
     adapter::{AdapterInfo, ModelAdapter},
     task::Task,
 };
-use crate::domain::tasks::{TextRecognitionConfig, TextRecognitionOutput, TextRecognitionTask};
+use crate::domain::tasks::{
+    TextRecognitionConfig, TextRecognitionOutput, TextRecognitionTask,
+    postprocess_text_direction_with_order_change,
+};
 use crate::impl_adapter_builder;
 use crate::models::recognition::crnn::{CRNNModel, CRNNModelBuilder, CRNNPreprocessConfig};
 
@@ -83,11 +86,18 @@ impl ModelAdapter for TextRecognitionAdapter {
             )
         {
             if score >= effective_config.score_threshold {
-                result_texts.push(text);
+                let (text, changed_order) = postprocess_text_direction_with_order_change(text);
                 result_scores.push(score);
-                result_positions.push(positions);
-                result_col_indices.push(col_indices);
-                result_seq_lengths.push(seq_len);
+                result_texts.push(text);
+                if changed_order {
+                    result_positions.push(Vec::new());
+                    result_col_indices.push(Vec::new());
+                    result_seq_lengths.push(0);
+                } else {
+                    result_positions.push(positions);
+                    result_col_indices.push(col_indices);
+                    result_seq_lengths.push(seq_len);
+                }
             } else {
                 // Keep entry to preserve index correspondence, but mark as filtered
                 result_texts.push(String::new());
@@ -179,6 +189,7 @@ impl_adapter_builder! {
             self.return_word_box = enable;
             self
         }
+
     }
 
     build: |builder: TextRecognitionAdapterBuilder, model_source: crate::core::ModelSource| -> Result<TextRecognitionAdapter, OCRError> {

oar-ocr-core/src/domain/tasks/mod.rs

@@ -44,4 +44,7 @@ pub use text_detection::{Detection, TextDetectionConfig, TextDetectionOutput, Te
 pub use text_line_orientation::{
     TextLineOrientationConfig, TextLineOrientationOutput, TextLineOrientationTask,
 };
-pub use text_recognition::{TextRecognitionConfig, TextRecognitionOutput, TextRecognitionTask};
+pub use text_recognition::{
+    TextRecognitionConfig, TextRecognitionOutput, TextRecognitionTask, postprocess_text_direction,
+    postprocess_text_direction_with_order_change,
+};

oar-ocr-core/src/domain/tasks/text_recognition.rs

@@ -9,6 +9,7 @@ use crate::core::traits::TaskDefinition;
 use crate::core::traits::task::{ImageTaskInput, Task, TaskSchema, TaskType};
 use crate::utils::ScoreValidator;
 use serde::{Deserialize, Serialize};
+use unicode_bidi::BidiInfo;
 
 /// Configuration for text recognition task.
 ///
@@ -28,6 +29,47 @@ impl Default for TextRecognitionConfig {
     }
 }
 
+fn reorder_bidi_line(line: &str) -> String {
+    let bidi_info = BidiInfo::new(line, None);
+    let Some(para) = bidi_info.paragraphs.first() else {
+        return line.to_string();
+    };
+
+    bidi_info
+        .reorder_line(para, para.range.clone())
+        .into_owned()
+}
+
+fn reorder_bidi_text(text: &str) -> (String, bool) {
+    let mut out = String::with_capacity(text.len());
+    let mut changed_order = false;
+
+    for segment in text.split_inclusive('\n') {
+        if let Some(line) = segment.strip_suffix('\n') {
+            let converted = reorder_bidi_line(line);
+            changed_order |= converted != line;
+            out.push_str(&converted);
+            out.push('\n');
+        } else {
+            let converted = reorder_bidi_line(segment);
+            changed_order |= converted != segment;
+            out.push_str(&converted);
+        }
+    }
+
+    (out, changed_order)
+}
+
+/// Applies Unicode bidi post-processing to decoded OCR text.
+pub fn postprocess_text_direction(text: String) -> String {
+    postprocess_text_direction_with_order_change(text).0
+}
+
+/// Applies Unicode bidi post-processing and reports whether character order changed.
+pub fn postprocess_text_direction_with_order_change(text: String) -> (String, bool) {
+    reorder_bidi_text(&text)
+}
+
 /// Output from text recognition task.
 #[derive(Debug, Clone)]
 pub struct TextRecognitionOutput {
@@ -185,4 +227,24 @@ mod tests {
         assert!(schema.input_types.contains(&"text_boxes".to_string()));
         assert!(schema.output_types.contains(&"text_strings".to_string()));
     }
+
+    #[test]
+    fn bidi_postprocess_leaves_ltr_text_unchanged() {
+        assert_eq!(
+            postprocess_text_direction("hello 123".to_string()),
+            "hello 123"
+        );
+    }
+
+    #[test]
+    fn bidi_postprocess_reports_order_changes() {
+        assert_eq!(
+            postprocess_text_direction_with_order_change("hello 123".to_string()),
+            ("hello 123".to_string(), false)
+        );
+        assert_eq!(
+            postprocess_text_direction_with_order_change("ابحرم".to_string()),
+            ("مرحبا".to_string(), true)
+        );
+    }
 }