Merge remote-tracking branch 'origin/main'

.gitignore

@@ -2,6 +2,10 @@ venv/
 .vscode/
 .cache/
 outputs/
+alto-outputs/
+page-outputs/
+text-outputs/
+tmp/
 
 # Byte-compiled / optimized / DLL files
 */__pycache__

app/main.py

@@ -1,20 +1,39 @@
 import gradio as gr
-
+import os
 from app.gradio_config import css, theme
-from app.tabs.submit import submit, custom_template_yaml
-from app.tabs.examples_tab import examples
+from app.tabs.submit import (
+    submit,
+    custom_template_yaml,
+    collection_submit_state,
+    batch_image_gallery,
+)
+from app.tabs.visualizer import visualizer, collection_viz_state, viz_image_gallery
 from app.tabs.templating import (
     templating_block,
     TEMPLATE_IMAGE_FOLDER,
     TEMPLATE_YAML_FOLDER,
     template_output_yaml_code,
 )
-from app.utils.md_helper import load_markdown
+
+from htrflow.models.huggingface.trocr import TrOCR
 
 gr.set_static_paths(paths=[TEMPLATE_IMAGE_FOLDER])
 gr.set_static_paths(paths=[TEMPLATE_YAML_FOLDER])
 
 
+def load_markdown(language, section, content_dir="app/content"):
+    """Load markdown content from files."""
+    if language is None:
+        file_path = os.path.join(content_dir, f"{section}.md")
+    else:
+        file_path = os.path.join(content_dir, language, f"{section}.md")
+
+    if os.path.exists(file_path):
+        with open(file_path, "r", encoding="utf-8") as f:
+            return f.read()
+    return f"## Content missing for {file_path} in {language}"
+
+
 with gr.Blocks(title="HTRflow", theme=theme, css=css) as demo:
     with gr.Row():
         with gr.Column(scale=1):
@@ -31,8 +50,8 @@ with gr.Blocks(title="HTRflow", theme=theme, css=css) as demo:
         with gr.Tab(label="Submit Job") as tab_submit:
             submit.render()
 
-        with gr.Tab(label="Output & Visualize") as tab_examples:
-            examples.render()
+        with gr.Tab(label="Visualize Result") as tab_visualizer:
+            visualizer.render()
 
     @demo.load(
         inputs=[template_output_yaml_code],
@@ -41,20 +60,41 @@ with gr.Blocks(title="HTRflow", theme=theme, css=css) as demo:
     def inital_yaml_code(template_output_yaml_code):
         return template_output_yaml_code
 
-    def sync_yaml_state(input_value, state_value):
+    @demo.load()
+    def inital_trocr_load():
+        return TrOCR("Riksarkivet/trocr-base-handwritten-hist-swe-2")
+
+    def sync_gradio_objects(input_value, state_value):
         """Synchronize the YAML state if there is a mismatch."""
         return input_value if input_value != state_value else gr.skip()
 
+    def sync_gradio_object_state(input_value, state_value):
+        """Synchronize the Collection."""
+        state_value = input_value
+        return state_value if state_value is not None else gr.skip()
+
+    tab_templating.select(
+        inputs=[custom_template_yaml, template_output_yaml_code],
+        outputs=[template_output_yaml_code],
+        fn=sync_gradio_objects,
+    )
+
     tab_submit.select(
         inputs=[template_output_yaml_code, custom_template_yaml],
         outputs=[custom_template_yaml],
-        fn=sync_yaml_state,
+        fn=sync_gradio_objects,
     )
 
-    tab_templating.select(
-        inputs=[custom_template_yaml, template_output_yaml_code],
-        outputs=[template_output_yaml_code],
-        fn=sync_yaml_state,
+    # tab_visualizer.select(
+    #     inputs=[batch_image_gallery, viz_image_gallery],
+    #     outputs=[viz_image_gallery],
+    #     fn=sync_gradio_objects,
+    # )
+
+    tab_visualizer.select(
+        inputs=[collection_submit_state, collection_viz_state],
+        outputs=[collection_viz_state],
+        fn=sync_gradio_object_state,
     )
 
 
@@ -63,7 +103,7 @@ demo.queue()
 if __name__ == "__main__":
     demo.launch(
         server_name="0.0.0.0",
-        server_port=7862,
+        server_port=7864,
         enable_monitoring=False,
-        show_error=True,
+        # show_error=True,
     )

app/tabs/examples_tab.py

@@ -1,89 +0,0 @@
-import gradio as gr
-from gradio_modal import Modal
-import pandas as pd
-
-
-output_image_placehholder = gr.Image(
-    label="Output image",
-    height=400,
-    show_share_button=True,
-)
-markdown_selected_option = gr.Markdown(value=" ", container=True)
-
-
-def htr_image_placehholder(txt, method, image):
-    needs_yaml_to_forward_tohtrflow_ = """steps:
-"""
-    print(method)
-
-    return txt, method, image
-
-
-with gr.Blocks() as examples:
-    with gr.Row(variant="panel"):
-        with gr.Column():
-            example_text_input_placeholder = gr.Markdown(visible=False, container=False)
-            example_method_input_placeholder = gr.Markdown(
-                visible=False, container=False
-            )
-            example_text_output_placeholder = gr.Markdown(
-                visible=False, container=False
-            )
-
-            user_image_input = gr.Image(visible=False, container=False)
-
-            # gr.Examples(  # TODO: add so example has a alto xml file on hf datasets
-            #     fn=htr_image_placehholder,
-            #     examples=images_for_demo.examples_list,
-            #     inputs=[
-            #         example_text_input_placeholder,
-            #         example_method_input_placeholder,
-            #         user_image_input,
-            #     ],
-            #     outputs=[example_text_output_placeholder, markdown_selected_option, output_image_placehholder],
-            #     cache_examples=True,
-            #     cache_mode="eager",
-            #     label="Example images",
-            #     examples_per_page=6,
-            # )
-
-        with gr.Column():
-            with gr.Tabs():
-                with gr.Tab("Viewer"):
-                    with gr.Group():
-                        with gr.Row():
-                            output_image_placehholder.render()
-                        with gr.Row():
-                            markdown_selected_option.render()
-                            show_output_modal = gr.Button(
-                                "Show output", variant="secondary", scale=0
-                            )
-                        with gr.Row():
-                            output_dataframe_pipeline = gr.Textbox(
-                                label="Click text", info="click on image bla bla.."
-                            )
-                with gr.Tab("YAML, DF, Graph?") as htrflow_output_table_tab:
-                    with gr.Group():
-                        with gr.Row():
-                            output_dataframe_pipeline = gr.Image(
-                                label="Output image", interactive=False, height="100"
-                            )
-                        with gr.Row():
-                            output_dataframe_pipeline = gr.Dataframe(
-                                label="Output image", col_count=2
-                            )
-
-                with Modal(visible=False) as output_modal:
-                    output_files_pipeline = gr.Files(
-                        label="Output files", height=100, visible=True, scale=0
-                    )
-                    output_yaml_code = gr.Code(
-                        language="yaml",
-                        label="yaml",
-                        interactive=True,
-                        visible=True,
-                        scale=0,
-                    )
-                    # TODO: yaml should parse from datasets.. and show the yaml code
-
-        show_output_modal.click(lambda: Modal(visible=True), None, output_modal)

app/tabs/submit.py

@@ -1,14 +1,152 @@
+import glob
 import time
+import uuid
 import gradio as gr
 from htrflow.pipeline.pipeline import Pipeline
+from htrflow.pipeline.steps import init_step
+import os
+from htrflow.volume.volume import Collection
+
 from htrflow.pipeline.steps import auto_import
 import yaml
 
+MAX_IMAGES = int(os.environ.get("MAX_IMAGES", 5))  # env: Maximum allowed images
+
+
+class PipelineWithProgress(Pipeline):
+    @classmethod
+    def from_config(cls, config: dict[str, str]):
+        """Init pipeline from config, ensuring the correct subclass is instantiated."""
+        return cls([init_step(step["step"], step.get("settings", {})) for step in config["steps"]])
+
+    def run(self, collection, start=0, progress=None):
+        """
+        Run pipeline on collection with Gradio progress support.
+        If progress is provided, it updates the Gradio progress bar during execution.
+        """
+        total_steps = len(self.steps[start:])
+        for i, step in enumerate(self.steps[start:]):
+            step_name = f"{step} (step {start + i + 1} / {total_steps})"
+
+            try:
+                progress((i + 1) / total_steps, desc=f"Running {step_name}")
+                collection = step.run(collection)
+
+            except Exception:
+                if self.pickle_path:
+                    gr.Error(
+                        f"HTRflow: Pipeline failed on step {step_name}. A backup collection is saved at {self.pickle_path}"
+                    )
+                else:
+                    gr.Error(
+                        f"HTRflow: Pipeline failed on step {step_name}",
+                    )
+                raise
+        return collection
+
+
+def rewrite_export_dests(config):
+    """
+    Rewrite the 'dest' in all 'Export' steps to include 'tmp' and a UUID.
+    Returns:
+        - A new config object with the updated 'dest' values.
+        - A list of all updated 'dest' paths.
+    """
+    new_config = {"steps": []}
+    updated_paths = []
+
+    unique_id = str(uuid.uuid4())
+
+    for step in config.get("steps", []):
+        new_step = step.copy()
+        if new_step.get("step") == "Export":
+            settings = new_step.get("settings", {})
+            if "dest" in settings:
+                new_dest = os.path.join("tmp", unique_id, settings["dest"])
+                settings["dest"] = new_dest
+                updated_paths.append(new_dest)
+        new_config["steps"].append(new_step)
+
+    return new_config, updated_paths
+
+
+def run_htrflow(custom_template_yaml, batch_image_gallery, progress=gr.Progress()):
+    """
+    Executes the HTRflow pipeline based on the provided YAML configuration and batch images.
+    Args:
+        custom_template_yaml (str): YAML string specifying the HTRflow pipeline configuration.
+        batch_image_gallery (list): List of uploaded images to process in the pipeline.
+    Returns:
+        tuple: A collection of processed items, list of exported file paths, and a Gradio update object.
+    """
+
+    if custom_template_yaml is None or len(custom_template_yaml) < 1:
+        gr.Warning("HTRflow: Please insert a HTRflow-yaml template")
+    try:
+        config = yaml.safe_load(custom_template_yaml)
+    except Exception as e:
+        gr.Warning(f"HTRflow: Error loading YAML configuration: {e}")
+        return gr.skip()
+
+    temp_config, tmp_output_paths = rewrite_export_dests(config)
+
+    progress(0, desc="HTRflow: Starting")
+    time.sleep(0.3)
 
-MAX_IMAGES = 5  # Maximum allowed images
+    if batch_image_gallery is None:
+        gr.Warning("HTRflow: You must upload atleast 1 image or more")
+
+    images = [temp_img[0] for temp_img in batch_image_gallery]
+
+    pipe = PipelineWithProgress.from_config(temp_config)
+    collections = auto_import(images)
+
+    gr.Info(f"HTRflow: processing {len(images)} {'image' if len(images) == 1 else 'images'}.")
+    progress(0.1, desc="HTRflow: Processing")
+
+    for collection in collections:
+        if "labels" in temp_config:
+            collection.set_label_format(**temp_config["labels"])
+
+        collection.label = "HTRflow_demo_output"
+        collection: Collection = pipe.run(collection, progress=progress)
+
+    exported_files = tracking_exported_files(tmp_output_paths)
+
+    time.sleep(0.5)
+    progress(1, desc="HTRflow: Finish")
+    gr.Info("HTRflow: Finish")
+
+    yield collection, exported_files, gr.skip()
+
+
+def tracking_exported_files(tmp_output_paths):
+    """
+    Look for files with specific extensions in the provided tmp_output_paths,
+    including subdirectories. Eliminates duplicate files.
+
+    Args:
+        tmp_output_paths (list): List of temporary output directories to search.
+
+    Returns:
+        list: Unique paths of all matching files found in the directories.
+    """
+    accepted_extensions = {".txt", ".xml", ".json"}
+
+    exported_files = set()
+
+    for tmp_folder in tmp_output_paths:
+        for ext in accepted_extensions:
+            search_pattern = os.path.join(tmp_folder, "**", f"*{ext}")
+            matching_files = glob.glob(search_pattern, recursive=True)
+            exported_files.update(matching_files)
+
+    return sorted(exported_files)
 
 
 with gr.Blocks() as submit:
+    collection_submit_state = gr.State()
+
     with gr.Column(variant="panel"):
         with gr.Group():
             with gr.Row():
@@ -20,6 +158,7 @@ with gr.Blocks() as submit:
                         height=400,
                         object_fit="cover",
                         columns=5,
+                        # preview=True,
                     )
 
                 with gr.Column(scale=1):
@@ -30,7 +169,9 @@ with gr.Blocks() as submit:
                         interactive=True,
                     )
         with gr.Row():
-            run_button = gr.Button("Submit", variant="primary", scale=0, min_width=160)
+            run_button = gr.Button("Submit", variant="primary", scale=0, min_width=200)
+            progess_bar = gr.Textbox(visible=False, show_label=False)
+            collection_output_files = gr.Files(label="Output Files", scale=0, min_width=400, visible=False)
 
     @batch_image_gallery.upload(
         inputs=batch_image_gallery,
@@ -42,38 +183,14 @@ with gr.Blocks() as submit:
             return gr.update(value=None)
         return images
 
-    def my_function(custom_template_yaml, batch_image_gallery, progress=gr.Progress()):
-        config = yaml.safe_load(custom_template_yaml)
-
-        image, _ = batch_image_gallery[0]
-
-        pipe = Pipeline.from_config(config)
-        print(batch_image_gallery)
-        collections = auto_import(image)
-
-        label = "HTRflow demo"
-
-        for collection in collections:
-            if "labels" in config:
-                collection.set_label_format(**config["labels"])
-            if label:
-                collection.label = label
-            collection = pipe.run(collection)
-
-        # progress(0, desc="Starting...")
-        # time.sleep(1)
-        # for i in progress.tqdm(range(100)):
-        #     if i == 20:
-        #         gr.Info("hej morgan")
-        #     if i == 50:
-        #         gr.Info("hej morgan2")
-        #     time.sleep(0.1)
-        # gr.Info("hej morgan nu är jag klar")
-        print(collection)
-        return gr.skip()
-
     run_button.click(
-        fn=my_function,
+        lambda: (gr.update(visible=True), gr.update(visible=False)),
+        outputs=[progess_bar, collection_output_files],
+    ).then(
+        fn=run_htrflow,
         inputs=[custom_template_yaml, batch_image_gallery],
-        outputs=batch_image_gallery,
+        outputs=[collection_submit_state, collection_output_files, progess_bar],
+    ).then(
+        lambda: (gr.update(visible=False), gr.update(visible=True)),
+        outputs=[progess_bar, collection_output_files],
     )

app/tabs/templating.py

@@ -6,11 +6,7 @@ import re
 def get_sorted_files(folder, extensions):
     """Retrieve sorted files by numeric value in their names."""
     return sorted(
-        [
-            os.path.join(folder, file)
-            for file in os.listdir(folder)
-            if file.lower().endswith(extensions)
-        ],
+        [os.path.join(folder, file) for file in os.listdir(folder) if file.lower().endswith(extensions)],
         key=lambda x: (
             int(re.search(r"\d+", os.path.basename(x)).group())
             if re.search(r"\d+", os.path.basename(x))
@@ -49,9 +45,7 @@ TEMPLATE_IMAGE_FOLDER = "app/assets/images"
 TEMPLATE_YAML_FOLDER = "app/assets/templates"
 
 # File Retrieval
-image_files = get_sorted_files(
-    TEMPLATE_IMAGE_FOLDER, (".png", ".jpg", ".jpeg", ".webp")
-)
+image_files = get_sorted_files(TEMPLATE_IMAGE_FOLDER, (".png", ".jpg", ".jpeg", ".webp"))
 yaml_files = get_sorted_files(TEMPLATE_YAML_FOLDER, (".yaml",))
 
 # Categorize YAML Files
@@ -96,11 +90,8 @@ with gr.Blocks() as templating_block:
             with gr.Group():
                 with gr.Row():
                     with gr.Column(scale=1):
-                        template_image = gr.Image(
-                            label="Example Templates", value=image_files[0], height=400
-                        )
+                        template_image = gr.Image(label="Example Templates", value=image_files[0], height=400)
                     with gr.Column(scale=1):
-
                         template_output_yaml_code = gr.Code(
                             language="yaml",
                             label="Pipeline",
@@ -108,7 +99,7 @@ with gr.Blocks() as templating_block:
                             visible=True,
                         )
     docs_link = gr.HTML(
-        value='<p><a href="https://ai-riksarkivet.github.io/htrflow/latest/getting_started/pipeline.html#example-pipelines" target="_blank">📚 Click here 📚</a> for a detailed description on how to customize the configuration</p>',
+        value='<p><a href="https://ai-riksarkivet.github.io/htrflow/latest/getting_started/pipeline.html#example-pipelines" target="_blank">📚 Click here 📚</a> for a detailed description on how to customize the configuration for HTRflow</p>',
         visible=True,
     )
 
@@ -121,7 +112,6 @@ with gr.Blocks() as templating_block:
         ],
     )
     def on_template_select(dropdown_selection_template):
-
         if dropdown_selection_template == "Simple":
             yaml_content = get_yaml_content(yaml_files_numbered[0])
             return image_files[0], yaml_content, gr.update(visible=False)
@@ -132,9 +122,7 @@ with gr.Blocks() as templating_block:
             yaml_content = get_yaml_content(yaml_files_c_letter[0])
             return image_files[2], yaml_content, gr.update(visible=True)
         else:
-            return gr.Error(
-                f"{dropdown_selection_template} - is not a valid Template selection"
-            )
+            return gr.Error(f"{dropdown_selection_template} - is not a valid Template selection")
 
     @custom_dropdown_selection_template.select(
         inputs=custom_dropdown_selection_template,
@@ -147,9 +135,7 @@ with gr.Blocks() as templating_block:
             yaml_content = get_yaml_content(yaml_path)
             return yaml_content
         else:
-            return gr.Error(
-                f"{custom_template_selection} - is not a valid Custom Template selection"
-            )
+            return gr.Error(f"{custom_template_selection} - is not a valid Custom Template selection")
 
     @dropdown_selection_template.select(
         inputs=dropdown_selection_template,

app/tabs/visualizer.py

@@ -0,0 +1,100 @@
+import gradio as gr
+import pandas as pd
+import numpy as np
+from htrflow.volume.volume import Collection
+from htrflow.utils.draw import draw_polygons
+from htrflow.utils import imgproc
+
+from htrflow.results import Segment
+
+
+with gr.Blocks() as visualizer:
+    with gr.Column(variant="panel"):
+        with gr.Row():
+            collection_viz_state = gr.State()
+            result_collection_viz_state = gr.State()
+            with gr.Column():
+                viz_image_gallery = gr.Gallery(
+                    file_types=["image"],
+                    label="Visualized Images from HTRflow",
+                    interactive=False,
+                    height=400,
+                    object_fit="cover",
+                    columns=5,
+                    preview=True,
+                )
+
+                visualize_button = gr.Button(
+                    "Visualize", scale=0, min_width=200, variant="secondary"
+                )
+
+            with gr.Column():
+                # image_visualizer_annotation = gr.Image(
+                #     interactive=False,
+                # )
+
+                line2 = gr.Gallery(
+                    interactive=False,
+                )
+                textlines = gr.Dataframe()
+
+    # @viz_image_gallery.select(outputs=image_visualizer_annotation)
+    # def return_image_from_gallery(evt: gr.SelectData):
+    #     return evt.value["image"]["path"]
+
+    @visualize_button.click(
+        outputs=[result_collection_viz_state, viz_image_gallery, line2, textlines]
+    )
+    def testie_load_pickle():
+        col = Collection.from_pickle(".cache/HTRflow_demo_output.pickle")
+
+        results = []
+        for page_idx, page_node in enumerate(col):
+            page_image = page_node.image.copy()
+
+            lines = list(page_node.traverse(lambda node: node.is_line()))
+
+            recog_conf_values = {
+                i: list(zip(tr.texts, tr.scores)) if (tr := ln.text_result) else []
+                for i, ln in enumerate(lines)
+            }
+
+            recog_df = pd.DataFrame(
+                [
+                    {"Transcription": text, "Confidence Score": f"{score:.4f}"}
+                    for values in recog_conf_values.values()
+                    for text, score in values
+                ]
+            )
+
+            line_polygons = []
+            line_crops = []
+            for ln in lines:
+                seg: Segment = ln.data.get("segment")
+                if not seg:
+                    continue
+
+                cropped_line_img = imgproc.crop(page_image, seg.bbox)
+                cropped_line_img = np.clip(cropped_line_img, 0, 255).astype(np.uint8)
+                line_crops.append(cropped_line_img)
+
+                if seg.polygon is not None:
+                    line_polygons.append(seg.polygon)
+
+            annotated_image = draw_polygons(page_image, line_polygons)
+            annotated_page_node = np.clip(annotated_image, 0, 255).astype(np.uint8)
+            results.append(
+                {
+                    "page_image": page_node,
+                    "annotated_page_node": annotated_page_node,
+                    "line_crops": line_crops,
+                    "recog_conf_values": recog_df,
+                }
+            )
+
+        return (
+            results,
+            [results[0]["annotated_page_node"]],
+            results[0]["line_crops"],
+            results[0]["recog_conf_values"],
+        )

app/utils/md_helper.py

@@ -1,14 +0,0 @@
-import os
-
-
-def load_markdown(language, section, content_dir="app/content"):
-    """Load markdown content from files."""
-    if language is None:
-        file_path = os.path.join(content_dir, f"{section}.md")
-    else:
-        file_path = os.path.join(content_dir, language, f"{section}.md")
-
-    if os.path.exists(file_path):
-        with open(file_path, "r", encoding="utf-8") as f:
-            return f.read()
-    return f"## Content missing for {file_path} in {language}"

app/utils/yaml_helper.py

@@ -1,91 +0,0 @@
-from jinja2 import Environment, FileSystemLoader
-
-
-def get_yaml_button_fn(
-    method,
-    output_formats,
-    reading_order,
-    simple_segment_model=None,
-    simple_htr_model=None,
-    simple_htr_model_type=None,
-    simple_segment_model_type=None,
-    nested_segment_model_1=None,
-    nested_segment_model_2=None,
-    nested_htr_model=None,
-    nested_segment_model_1_type=None,
-    nested_segment_model_2_type=None,
-    nested_htr_model_type=None,
-):
-    env = Environment(loader=FileSystemLoader("app/templates"))
-
-    if output_formats is None:
-        output_formats = ["txt"]
-
-    template_name = "steps_template.yaml.j2"
-    try:
-        if method == "Simple layout":
-            steps = [
-                {
-                    "step": "Segmentation",
-                    "model": simple_segment_model_type,
-                    "model_settings": {"model": simple_segment_model},
-                },
-                {
-                    "step": "TextRecognition",
-                    "model": simple_htr_model_type,
-                    "model_settings": {"model": simple_htr_model},
-                },
-            ]
-        elif method == "Nested segmentation":
-            steps = [
-                {
-                    "step": "Segmentation",
-                    "model": nested_segment_model_1_type,
-                    "model_settings": {"model": nested_segment_model_1},
-                },
-                {
-                    "step": "Segmentation",
-                    "model": nested_segment_model_2_type,
-                    "model_settings": {"model": nested_segment_model_2},
-                },
-                {
-                    "step": "TextRecognition",
-                    "model": nested_htr_model_type,
-                    "model_settings": {"model": nested_htr_model},
-                },
-            ]
-        else:
-            return "Invalid method or not yet supported."
-
-        steps.append({"step": reading_order})
-
-        # TODO: fix reading order
-        # - step: ReadingOrderMarginalia
-        # settings:
-        #     two_page: always
-
-        # TODO: fix labeling format
-        # # Sets label format to regionX_lineY_wordZ
-        # labels:
-        # level_labels:
-        #     - region
-        #     - line
-        #     - word
-        # sep: _
-        # template: "{label}{number}"
-
-        steps.extend(
-            {
-                "step": "Export",
-                "settings": {"format": format, "dest": f"{format}-outputs"},
-            }
-            for format in output_formats
-        )
-
-        template = env.get_template(template_name)
-
-        yaml_value = template.render(steps=steps)
-        return yaml_value
-
-    except Exception as e:
-        return f"Error generating YAML: {str(e)}"

pyproject.toml

@@ -17,13 +17,12 @@ classifiers = [
 requires-python = ">=3.10,<3.13"
 
 dependencies = [
-    "htrflow==0.1.3",
+    "htrflow==0.2.0",
     "gradio>=5.11.0",
     "datasets>=3.2.0",
     "pandas>=2.2.3",
-    "jinja2>=3.1.4",
-    "gradio-modal>=0.0.4",
-    "fastapi[standard]>=0.115.6",
+    "tqdm>=4.67.1",
+    "pillow>=11.1.0",
 ]
 
 [project.urls]
@@ -46,6 +45,9 @@ openmmlab = [
     "mmdet==3.0.0",
     "mmocr==1.0.0"
 ]
+teklia = [
+    "pylaia==1.1.2"
+]
 
 
 [build-system]

uv.lock

@@ -851,18 +851,6 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/a6/3d/e05202dd42581c2a1e93c730d10a0ef45bc40921332c9aa0d6645bbf0e2b/gradio_client-1.5.4-py3-none-any.whl", hash = "sha256:ad38c9a6f7fc590e822627f5bf5685321a7822b8f1a88b76d00a0621a43162d6", size = 321364 },
 ]
 
-[[package]]
-name = "gradio-modal"
-version = "0.0.4"
-source = { registry = "https://pypi.org/simple" }
-dependencies = [
-    { name = "gradio" },
-]
-sdist = { url = "https://files.pythonhosted.org/packages/e2/fd/3b383f9ee8d60625e9e26871ba4adcacbedeab132041b94290758e02e543/gradio_modal-0.0.4.tar.gz", hash = "sha256:717ae699072a171648cfa1b84bc153be84e92d04e9ad58c1bc59af68ef332726", size = 1180812 }
-wheels = [
-    { url = "https://files.pythonhosted.org/packages/05/3d/76f454de84ae1dccbf2b7023e933afb8dde5fdd89e9476786726ef770737/gradio_modal-0.0.4-py3-none-any.whl", hash = "sha256:d96e817d2e934d9e1b835b06474f45fd349b5ccea499d1536bfb4bd38f62dedb", size = 1106241 },
-]
-
 [[package]]
 name = "h11"
 version = "0.14.0"
@@ -905,10 +893,10 @@ dependencies = [
     { name = "datasets" },
     { name = "fastapi", extra = ["standard"] },
     { name = "gradio" },
-    { name = "gradio-modal" },
     { name = "htrflow" },
-    { name = "jinja2" },
     { name = "pandas" },
+    { name = "pillow" },
+    { name = "tqdm" },
 ]
 
 [package.optional-dependencies]
@@ -933,15 +921,15 @@ requires-dist = [
     { name = "datasets", specifier = ">=3.2.0" },
     { name = "fastapi", extras = ["standard"], specifier = ">=0.115.6" },
     { name = "gradio", specifier = ">=5.11.0" },
-    { name = "gradio-modal", specifier = ">=0.0.4" },
     { name = "htrflow", specifier = "==0.1.3" },
-    { name = "jinja2", specifier = ">=3.1.4" },
     { name = "mmcv", marker = "extra == 'openmmlab'", specifier = "==2.0.1" },
     { name = "mmdet", marker = "extra == 'openmmlab'", specifier = "==3.0.0" },
     { name = "mmengine", marker = "extra == 'openmmlab'", specifier = "==0.7.4" },
     { name = "mmocr", marker = "extra == 'openmmlab'", specifier = "==1.0.0" },
     { name = "openmim", marker = "extra == 'openmmlab'", specifier = "==0.3.9" },
     { name = "pandas", specifier = ">=2.2.3" },
+    { name = "pillow", specifier = ">=11.1.0" },
+    { name = "tqdm", specifier = ">=4.67.1" },
 ]
 
 [package.metadata.requires-dev]