fixed spaces zerogpu

README.md

@@ -3,9 +3,13 @@ title: HTRFLOW
 emoji: 🏢
 colorFrom: purple
 colorTo: green
-sdk: docker
+sdk: gradio
+app_file: app/main.py
 pinned: false
 license: mit
 short_description: htrflow demo app
 header: mini
----
+thumbnail: >-
+  https://cdn-uploads.huggingface.co/production/uploads/60a4e677917119d38f6bbff8/-qMf3PaegicobqW5hXyiA.png
+sdk_version: 5.17.0
+---

app/assets/templates/nested_swe_ra.yaml

@@ -4,18 +4,22 @@ steps:
     model: yolo
     model_settings:
        model: Riksarkivet/yolov9-regions-1
+    generation_settings:
+       batch_size: 4
 - step: Segmentation
   settings:
     model: yolo
     model_settings:
       model: Riksarkivet/yolov9-lines-within-regions-1
+    generation_settings:
+       batch_size: 8
 - step: TextRecognition
   settings:
     model: TrOCR
     model_settings:
       model: Riksarkivet/trocr-base-handwritten-hist-swe-2
     generation_settings:
-       batch_size: 4
+       batch_size: 16
 - step: ReadingOrderMarginalia
   settings:
     two_page: True

app/assets/templates/simple_eng_modern.yaml

@@ -9,4 +9,6 @@ steps:
     model: TrOCR
     model_settings:
       model: microsoft/trocr-base-handwritten
+    generation_settings:
+       batch_size: 16
 - step: OrderLines

app/assets/templates/simple_medival.yaml

@@ -9,4 +9,6 @@ steps:
     model: TrOCR
     model_settings:
       model: medieval-data/trocr-medieval-base
+    generation_settings:
+       batch_size: 16
 - step: OrderLines

app/assets/templates/simple_nordhand.yaml

@@ -9,4 +9,6 @@ steps:
     model: TrOCR
     model_settings:
       model: Sprakbanken/TrOCR-norhand-v3
+    generation_settings:
+       batch_size: 16
 - step: OrderLines

app/assets/templates/simple_pylaia_french.yaml

@@ -1,12 +0,0 @@
-steps:
-- step: Segmentation
-  settings:
-    model: yolo
-    model_settings:
-      model: Riksarkivet/yolov9-lines-within-regions-1
-- step: TextRecognition
-  settings:
-    model: Pylaia
-    model_settings:
-      model: Teklia/pylaia-belfort
-- step: OrderLines

app/assets/templates/simple_swe_ra.yaml

@@ -9,4 +9,6 @@ steps:
     model: TrOCR
     model_settings:
       model: Riksarkivet/trocr-base-handwritten-hist-swe-2
+    generation_settings:
+       batch_size: 16
 - step: OrderLines

app/main.py

@@ -113,5 +113,5 @@ demo.queue()
 
 if __name__ == "__main__":
     demo.launch(
-        server_name="0.0.0.0", server_port=7860, enable_monitoring=True, show_api=False
+        server_name="0.0.0.0", server_port=7860, enable_monitoring=False, show_api=False
     )

app/pipelines.py

@@ -23,11 +23,6 @@ PIPELINES = {
         "description": "This pipeline works well on handwritten historic letters and other documents written in Norwegian with only one text region. The model is developed by the <a href='https://huggingface.co/Sprakbanken/TrOCR-norhand-v3'>Language Bank</a> at The National Library of Norway.",
         "examples": ["norhand_fmgh040_4.jpg"],
     },
-    "French - Single page and snippets": {
-        "file": "app/assets/templates/simple_pylaia_french.yaml",
-        "description": "This pipeline works well on handwritten historic letters and other documents written in French. The model is developed by <a href='https://huggingface.co/Teklia/pylaia-belfort'>Teklia</a>.",
-        "examples": ["belfort_snippet.jpg"],
-    },
     "Medival - Single page and snippets": {
         "file": "app/assets/templates/simple_medival.yaml",
         "description": "This pipeline works well for medieval scripts written in single-page running text. It is a base model from <a href='https://huggingface.co/medieval-data'>Medieval Data</a>, but other models can be selected from here: <a href='https://huggingface.co/collections/medieval-data/trocr-medieval-htr-66871faba03abfbb1b66ab69'>Medieval Models</a>.",

app/tabs/submit.py

@@ -3,6 +3,7 @@ import os
 import time
 
 import gradio as gr
+import spaces
 import yaml
 from gradio_modal import Modal
 from htrflow.pipeline.pipeline import Pipeline
@@ -64,6 +65,7 @@ class PipelineWithProgress(Pipeline):
         return collection
 
 
+@spaces.GPU
 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.
@@ -105,7 +107,7 @@ def run_htrflow(custom_template_yaml, batch_image_gallery, progress=gr.Progress(
 
     progress(1, desc="HTRflow: Finish, redirecting to 'Results tab'")
     time.sleep(2)
-    gr.Info("Image were succesfully transcribed ✨")
+    gr.Info("Completed succesfully ✨")
 
     yield collection, gr.skip()
 
@@ -156,11 +158,24 @@ def get_selected_example_pipeline(event: gr.SelectData) -> str | None:
             return name
 
 
-def get_image_from_image_id(image_id):
+def get_image_from_image_url(input_value):
     """
-    Get URL of submitted image ID.
+    Get URL of the image from either an image_id (from Riksarkivet) or an image_url directly.
+    If input_value is an image_id, it constructs the IIIF URL.
+    If input_value is an image_url, it returns the URL as-is.
     """
-    return [f"https://lbiiif.riksarkivet.se/arkis!{image_id}/full/max/0/default.jpg"]
+
+    if input_value.startswith("http"):
+        return [input_value]
+    else:
+        input_value = input_value.split(",")
+
+        return [
+            (
+                f"https://lbiiif.riksarkivet.se/arkis!{item.strip()}/full/max/0/default.jpg"
+            )
+            for item in input_value
+        ]
 
 
 with gr.Blocks() as submit:
@@ -279,17 +294,8 @@ with gr.Blocks() as submit:
             return gr.update(value=None)
         return images
 
-    def return_iiif_url(image_ids):
-        if isinstance(image_ids, str):
-            image_ids = image_ids.split(",")
-
-        return [
-            f"https://lbiiif.riksarkivet.se/arkis!{image_id.strip()}/full/max/0/default.jpg"
-            for image_id in image_ids
-        ]
-
     image_iiif_url.submit(
-        fn=return_iiif_url, inputs=image_iiif_url, outputs=batch_image_gallery
+        fn=get_image_from_image_url, inputs=image_iiif_url, outputs=batch_image_gallery
     ).then(fn=lambda: "Swedish - Spreads", outputs=pipeline_dropdown)
 
     run_button.click(
@@ -302,5 +308,3 @@ with gr.Blocks() as submit:
     examples.select(get_selected_example_pipeline, None, pipeline_dropdown)
 
     edit_pipeline_button.click(lambda: Modal(visible=True), None, edit_pipeline_modal)
-
-# TODO: submit on image_id .. --> yaml swedish - spreads

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "htrflow-app"
-version = "1.0.0"
+version = "1.0.1"
 description = "htrflow-app is developed at Riksarkivet's AI-lab as a poc frontend to showcase HTRflow"
 readme = "README.md"
 license = {file = "LICENSE"}
@@ -18,11 +18,11 @@ requires-python = ">=3.10,<3.13"
 
 dependencies = [
     "htrflow==0.2.5",
-    "gradio>=5.16.1",
+    "gradio>=5.17.0",
     "tqdm>=4.67.1",
     "gradio-modal>=0.0.4",
     "dill>=0.3.9",
-    "pylaia>=1.1.2",
+    "spaces>=0.32.0",
 ]
 
 [project.urls]

requirements.txt

@@ -0,0 +1,5 @@
+htrflow==0.2.5
+gradio>=5.17.0
+tqdm>=4.67.1
+gradio-modal>=0.0.4
+dill>=0.3.9