Table parser

.gitattributes

@@ -0,0 +1 @@
+*.onnx filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -6,6 +6,8 @@ model-archive/
 *.pt
 test_data/*
 
-models/
+models/*
+!models/table-transformer-structure-recognition_fp16.onnx
 
 **/.env
+**/examples/

ferrules-cli/src/main.rs

@@ -229,6 +229,9 @@ fn parse_ep_args(args: &Args) -> Vec<OrtExecutionProvider> {
 #[tokio::main(flavor = "multi_thread")]
 async fn main() {
     let args = Args::parse();
+    if args.debug || std::env::var("RUST_LOG").is_ok() {
+        tracing_subscriber::fmt::init();
+    }
 
     // Check providers
     let providers = parse_ep_args(&args);
@@ -239,8 +242,6 @@ async fn main() {
         inter_threads: args.inter_threads,
         opt_level: args.graph_opt_level.map(|v| v.try_into().unwrap()),
     };
-    // Global tasks
-    let parser = FerrulesParser::new(ort_config);
 
     let page_range = match args.page_range {
         Some(ref page_range_str) => match parse_page_range(page_range_str) {
@@ -263,10 +264,12 @@ async fn main() {
         },
         None => None,
     };
-
     let pb = setup_progress_bar(&args.file_path, None, page_range.clone());
     let pbc = pb.clone();
 
+    // Global tasks
+    let parser = FerrulesParser::new(ort_config);
+
     let doc_name = args
         .file_path
         .file_name()
@@ -440,6 +443,20 @@ async fn main() {
                         ],
                     );
                 }
+                ferrules_core::error::FerrulesError::TableTransformerModelError(e) => {
+                    format_error(
+                        "Table Transformation Failed",
+                        "Failed to process table using the vision model.",
+                        vec![
+                            ("Error", e),
+                            ("File", args.file_path.display().to_string()),
+                            (
+                                "Suggestion",
+                                "Check if the model files are present and valid.".to_string(),
+                            ),
+                        ],
+                    );
+                }
             }
             std::process::exit(1);
         }

ferrules-core/Cargo.toml

@@ -10,6 +10,7 @@ itertools = "0.14.0"
 futures = "0.3.31"
 colored = "3.0.0"
 dirs = "6.0.0"
+half = "2.4.1"
 anyhow = { workspace = true }
 uuid = { workspace = true }
 tracing = { workspace = true }
@@ -39,7 +40,7 @@ regex = "1.11.1"
 html2md = "0.2.15"
 
 [target.'cfg(target_os = "macos")'.dependencies]
-ort = { version = "=2.0.0-rc.9", features = ["coreml", "fetch-models"] }
+ort = { version = "=2.0.0-rc.9", features = ["coreml", "fetch-models", "half"] }
 objc2 = { version = "^0.5.2" }
 objc2-foundation = { version = "^0.2.2" }
 objc2-vision = { version = "^0.2.2", features = [

ferrules-core/src/blocks.rs

@@ -28,6 +28,46 @@ pub struct List {
     pub(crate) items: Vec<String>,
 }
 
+#[derive(Clone, Debug, Default, Deserialize, Serialize)]
+pub enum TableAlgorithm {
+    #[default]
+    Unknown,
+    Lattice,
+    Stream,
+    Vision,
+}
+
+#[derive(Clone, Debug, Default, Deserialize, Serialize)]
+pub struct TableBlock {
+    pub(crate) id: usize,
+    pub(crate) caption: Option<String>,
+    pub rows: Vec<TableRow>,
+    pub has_borders: bool,
+    pub algorithm: TableAlgorithm,
+}
+
+impl TableBlock {
+    pub(crate) fn path(&self) -> String {
+        format!("table_{}.png", self.id)
+    }
+}
+
+#[derive(Clone, Debug, Default, Deserialize, Serialize)]
+pub struct TableRow {
+    pub cells: Vec<TableCell>,
+    pub is_header: bool,
+    pub bbox: BBox,
+}
+
+#[derive(Clone, Debug, Default, Deserialize, Serialize)]
+pub struct TableCell {
+    pub content: Vec<Block>,
+    pub text: String,
+    pub row_span: u8,
+    pub col_span: u8,
+    pub bbox: BBox,
+}
+
 #[derive(Clone, Debug, Default, Deserialize, Serialize)]
 pub struct Title {
     pub level: TitleLevel,
@@ -43,7 +83,7 @@ pub enum BlockType {
     ListBlock(List),
     TextBlock(TextBlock),
     Image(ImageBlock),
-    Table,
+    Table(TableBlock),
 }
 
 impl std::fmt::Display for BlockType {
@@ -121,7 +161,21 @@ impl Block {
             }
             BlockType::Title(_title) => todo!(),
             BlockType::Image(_image_block) => todo!(),
-            BlockType::Table => todo!(),
+            BlockType::Table(table) => {
+                if let ElementType::Table(incoming_table_opt) = &element.kind {
+                    self.bbox.merge(&element.bbox);
+                    if let Some(incoming_table) = incoming_table_opt {
+                        table.rows.extend(incoming_table.rows.clone());
+                    }
+                    Ok(())
+                } else {
+                    Err(FerrulesError::BlockMergeError {
+                        element,
+                        block_id: self.id,
+                        kind: self.kind.clone(),
+                    })
+                }
+            }
         }
     }
 
@@ -133,7 +187,7 @@ impl Block {
             BlockType::Title(_) => "TITLE",
             BlockType::ListBlock(_) => "LIST",
             BlockType::Image(_) => "IMAGE",
-            BlockType::Table => "TABLE",
+            BlockType::Table(_) => "TABLE",
         }
     }
 }