Fixed some texts and added PDF upload functionality

app/content/sidebar.md

@@ -1,19 +1,19 @@
 ## HTRflow Demo
-
-Welcome to the **HTRflow Demo** – a web application developed by the **National Archives of Sweden** in collaboration with [Huminfra](https://www.huminfra.se/). This demo lets you explore how AI transforms historical manuscripts into digital text using [HTRflow](https://ai-riksarkivet.github.io/htrflow/latest).
+Developed by the **National Archives of Sweden** with [Huminfra](https://www.huminfra.se/) that demonstrates AI-powered conversion of historical manuscripts to digital text using [HTRflow](https://ai-riksarkivet.github.io/htrflow/latest).
 
 > Note: This demo application is for demonstration purposes only and is not intended for production use.
 > The application is hosted on Hugging Face 🤗 using shared infrastructure, which means there is a daily quota limit on how much you can use the app each day.
 
-### Contact
-
-✉️ [[email protected]](mailto:[email protected])
-
-### Projects
-
 Both the App and HTRflow's code are completely open source. Explore and contribute on GitHub:
 
 - [APP](https://github.com/AI-Riksarkivet/htrflow_app).
 - [HTRflow](https://ai-riksarkivet.github.io/htrflow/latest).
 
 If you find our projects useful, please consider giving us a star ⭐!
+
+
+### Contact
+
+✉️ [[email protected]](mailto:[email protected])
+
+

app/pipelines.py

@@ -1,7 +1,7 @@
 PIPELINES = {
     "Swedish - Spreads": {
         "file": "app/assets/templates/nested_swe_ra.yaml",
-        "description": "This pipeline works well on handwritten historic documents written in Swedish with multiple text regions. The model is devloped by <a href='https://huggingface.co/Riksarkivet'>the National Archives of Sweden</a>.",
+        "description": "This pipeline works well on handwritten historic documents written in Swedish with multiple text regions. The HTR model used in the pipeline is <b>Swedish Lion Libre</b> from <a href='https://huggingface.co/Riksarkivet'>the National Archives of Sweden</a>.",
         "examples": [
             "R0003364_00005.jpg",
             "30002027_00008.jpg",
@@ -10,7 +10,7 @@ PIPELINES = {
     },
     "Swedish - Single page and snippets": {
         "file": "app/assets/templates/simple_swe_ra.yaml",
-        "description": "This pipeline works well on handwritten historic letters and other documents written in Swedish with only one text region. The model is devloped by <a href='https://huggingface.co/Riksarkivet'>the National Archives of Sweden</a>.",
+        "description": "This pipeline works well on handwritten historic letters and other documents written in Swedish with only one text region. The HTR model used in the pipeline is <b>Swedish Lion Libre</b> from <a href='https://huggingface.co/Riksarkivet'>the National Archives of Sweden</a>.",
         "examples": [
             "451511_1512_01.jpg",
             "A0062408_00006.jpg",
@@ -23,14 +23,14 @@ 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"],
     },
-    "Medival - Single page and snippets": {
+    "Medieval - 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>.",
+        "description": "This pipeline works well for medieval scripts written in single-page running text. The HTR model is 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>.",
         "examples": ["manusscript_kb.png"],
     },
     "English - Single page and snippets": {
         "file": "app/assets/templates/simple_eng_modern.yaml",
-        "description": "This pipeline works well for English text in single page running text. This a base model from <a href='https://huggingface.co/microsoft/trocr-base-handwritten'>Microsoft</a>.",
+        "description": "This pipeline works well for English text in single page running text. The HTR model is from <a href='https://huggingface.co/microsoft/trocr-base-handwritten'>Microsoft</a>.",
         "examples": ["iam.png"],
     },
 }

app/tabs/submit.py

@@ -13,6 +13,7 @@ from gradio_modal import Modal
 from htrflow.pipeline.pipeline import Pipeline
 from htrflow.pipeline.steps import init_step
 from htrflow.volume.volume import Collection
+from pdf2image import convert_from_path
 
 from app.pipelines import PIPELINES
 
@@ -36,7 +37,12 @@ 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"]])
+        return cls(
+            [
+                init_step(step["step"], step.get("settings", {}))
+                for step in config["steps"]
+            ]
+        )
 
     def run(self, collection, start=0, progress=None):
         """
@@ -95,7 +101,9 @@ def run_htrflow(custom_template_yaml, batch_image_gallery, progress=gr.Progress(
 
     pipe = PipelineWithProgress.from_config(config)
 
-    gr.Info(f"HTRflow: processing {len(images)} {'image' if len(images) == 1 else 'images'}.")
+    gr.Info(
+        f"HTRflow: processing {len(images)} {'image' if len(images) == 1 else 'images'}."
+    )
     progress(0.1, desc="HTRflow: Processing")
 
     collection.label = "demo_output"
@@ -192,7 +200,7 @@ def get_image_from_image_id(image_id):
 def get_images_from_iiif_manifest(iiif_manifest_url, max_images=20, height=1200):
     """
     Read images from a v2/v3 IIIF manifest, limited to max_images.
-    
+
     Arguments:
         iiif_manifest_url: URL to IIIF manifest
         height: Max height of returned images
@@ -201,7 +209,7 @@ def get_images_from_iiif_manifest(iiif_manifest_url, max_images=20, height=1200)
     try:
         buffer = io.BytesIO()
         c = pycurl.Curl()
-        
+
         c.setopt(c.URL, iiif_manifest_url)
         c.setopt(c.WRITEDATA, buffer)
         c.setopt(c.CAINFO, certifi.where())
@@ -211,42 +219,48 @@ def get_images_from_iiif_manifest(iiif_manifest_url, max_images=20, height=1200)
         c.setopt(c.TIMEOUT, 10)
         c.setopt(c.NOSIGNAL, 1)
         c.setopt(c.USERAGENT, "curl/7.68.0")
-        
+
         c.perform()
-        
+
         http_code = c.getinfo(c.RESPONSE_CODE)
         if http_code != 200:
             raise Exception(f"HTTP Error: {http_code}")
-        
+
         manifest = buffer.getvalue().decode("utf-8")
         c.close()
-        
+
     except pycurl.error as e:
         error_code, error_msg = e.args
-        raise Exception(f"Could not fetch IIIF manifest from {iiif_manifest_url} ({error_msg})")
-    
+        raise Exception(
+            f"Could not fetch IIIF manifest from {iiif_manifest_url} ({error_msg})"
+        )
+
     # Hacky solution to get all images regardless of API version - treat
     # the manifest as a string and match everything that looks like an IIIF
     # image URL.
     pattern = r'(?P<identifier>https?://[^"\s]*)/(?P<region>[^"\s]*?)/(?P<size>[^"\s]*?)/(?P<rotation>!?\d*?)/(?P<quality>[^"\s]*?)\.(?P<format>jpg|tif|png|gif|jp2|pdf|webp)'
-    
-    images = set()  # create a set to eliminate duplicates (e.g. thumbnails and fullsize images)
-    
+
+    images = (
+        set()
+    )  # create a set to eliminate duplicates (e.g. thumbnails and fullsize images)
+
     for match in re.findall(pattern, manifest):
         identifier, _, _, _, _, format_ = match
         images.add(f"{identifier}/full/{height},/0/default.{format_}")
-        
+
         # Stop adding images if we've reached the maximum
         if len(images) >= max_images:
             break
-    
+
     # Sort and limit the results to max_images
     return sorted(images)[:max_images], gr.update(visible=True)
 
 
 with gr.Blocks() as submit:
     gr.Markdown("# Upload")
-    gr.Markdown("Select or upload the image you want to transcribe. You can upload up to five images at a time.")
+    gr.Markdown(
+        "Select or upload the image you want to transcribe. Most common image formats are supported and you can upload max 5 images at a time in this hosted demo."
+    )
 
     collection_submit_state = gr.State()
 
@@ -293,11 +307,16 @@ with gr.Blocks() as submit:
                                 "Use an image from a IIIF manifest by pasting a IIIF manifest URL. Press enter to submit."
                             ),
                             placeholder="",
-                            scale=0
+                            scale=0,
                         )
-                        max_images_iiif_manifest= gr.Number(value=20, min_width=50, scale=0,
+                        max_images_iiif_manifest = gr.Number(
+                            value=20,
+                            min_width=50,
+                            scale=0,
                             label="Number of image to return from IIIF manifest",
-                            minimum=1, visible=False)
+                            minimum=1,
+                            visible=False,
+                        )
                     iiif_gallery = gr.Gallery(
                         interactive=False,
                         columns=4,
@@ -314,6 +333,18 @@ with gr.Blocks() as submit:
                         placeholder="https://example.com/image.jpg",
                     )
 
+                with gr.Tab("PDF"):
+                    pdf_file = gr.File(label="PDF", file_types=[".pdf"])
+
+                    pdf_gallery = gr.Gallery(
+                        interactive=False,
+                        columns=4,
+                        allow_preview=False,
+                        container=False,
+                        show_label=False,
+                        object_fit="scale-down",
+                    )
+
     with gr.Column(variant="panel", elem_classes="panel-with-border"):
         gr.Markdown("## Settings")
         gr.Markdown(
@@ -370,7 +401,7 @@ with gr.Blocks() as submit:
         )
 
     with gr.Row():
-        run_button = gr.Button("Transcribe", variant="primary", scale=0, min_width=200)
+        run_button = gr.Button("Run HTR", variant="primary", scale=0, min_width=200)
 
     @batch_image_gallery.upload(
         inputs=batch_image_gallery,
@@ -385,9 +416,17 @@ with gr.Blocks() as submit:
     image_id.submit(get_image_from_image_id, image_id, batch_image_gallery).then(
         fn=lambda: "Swedish - Spreads", outputs=pipeline_dropdown
     )
-    iiif_manifest_url.submit(get_images_from_iiif_manifest, [iiif_manifest_url, max_images_iiif_manifest], [iiif_gallery, max_images_iiif_manifest])
+    iiif_manifest_url.submit(
+        get_images_from_iiif_manifest,
+        [iiif_manifest_url, max_images_iiif_manifest],
+        [iiif_gallery, max_images_iiif_manifest],
+    )
     image_url.submit(lambda url: [url], image_url, batch_image_gallery)
 
+    pdf_file.upload(
+        lambda imgs: convert_from_path(imgs), inputs=pdf_file, outputs=pdf_gallery
+    )
+
     run_button.click(
         fn=run_htrflow,
         inputs=[custom_template_yaml, batch_image_gallery],
@@ -398,5 +437,6 @@ with gr.Blocks() as submit:
     examples.select(get_selected_example_pipeline, None, pipeline_dropdown)
 
     iiif_gallery.select(get_selected_example_image, None, batch_image_gallery)
+    pdf_gallery.select(get_selected_example_image, None, batch_image_gallery)
 
     edit_pipeline_button.click(lambda: Modal(visible=True), None, edit_pipeline_modal)

pyproject.toml

@@ -24,6 +24,7 @@ dependencies = [
     "dill>=0.3.9",
     "spaces>=0.32.0",
     "pycurl",
+    "pdf2image>=1.17.0",
 ]
 
 [project.urls]

requirements.txt

@@ -3,4 +3,5 @@ gradio>=5.20.1
 tqdm>=4.67.1
 gradio-modal>=0.0.4
 dill>=0.3.9
-pycurl>=7.45.6
+pycurl>=7.45.6
+pdf2image>=1.17.0

uv.lock

@@ -543,6 +543,8 @@ dependencies = [
     { name = "gradio" },
     { name = "gradio-modal" },
     { name = "htrflow" },
+    { name = "pdf2image" },
+    { name = "pycurl" },
     { name = "spaces" },
     { name = "tqdm" },
 ]
@@ -560,6 +562,8 @@ requires-dist = [
     { name = "gradio", specifier = ">=5.17.0" },
     { name = "gradio-modal", specifier = ">=0.0.4" },
     { name = "htrflow", specifier = "==0.2.5" },
+    { name = "pdf2image", specifier = ">=1.17.0" },
+    { name = "pycurl" },
     { name = "spaces", specifier = ">=0.32.0" },
     { name = "tqdm", specifier = ">=4.67.1" },
 ]
@@ -1181,6 +1185,18 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/29/d4/1244ab8edf173a10fd601f7e13b9566c1b525c4f365d6bee918e68381889/pandas-2.2.3-cp312-cp312-win_amd64.whl", hash = "sha256:59ef3764d0fe818125a5097d2ae867ca3fa64df032331b7e0917cf5d7bf66b13", size = 11504248 },
 ]
 
+[[package]]
+name = "pdf2image"
+version = "1.17.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "pillow" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/00/d8/b280f01045555dc257b8153c00dee3bc75830f91a744cd5f84ef3a0a64b1/pdf2image-1.17.0.tar.gz", hash = "sha256:eaa959bc116b420dd7ec415fcae49b98100dda3dd18cd2fdfa86d09f112f6d57", size = 12811 }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/62/33/61766ae033518957f877ab246f87ca30a85b778ebaad65b7f74fa7e52988/pdf2image-1.17.0-py3-none-any.whl", hash = "sha256:ecdd58d7afb810dffe21ef2b1bbc057ef434dabbac6c33778a38a3f7744a27e2", size = 11618 },
+]
+
 [[package]]
 name = "pfzy"
 version = "0.3.4"
@@ -1357,6 +1373,32 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/2d/c7/a0d3356f3074ac548afefa515ff46f3bea011deca607faf1c09b26dd5330/pycryptodomex-3.21.0-pp310-pypy310_pp73-win_amd64.whl", hash = "sha256:27e84eeff24250ffec32722334749ac2a57a5fd60332cd6a0680090e7c42877e", size = 1792099 },
 ]
 
+[[package]]
+name = "pycurl"
+version = "7.45.6"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/71/35/fe5088d914905391ef2995102cf5e1892cf32cab1fa6ef8130631c89ec01/pycurl-7.45.6.tar.gz", hash = "sha256:2b73e66b22719ea48ac08a93fc88e57ef36d46d03cb09d972063c9aa86bb74e6", size = 239470 }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/31/11/a491677a902053e2720c29e9bc743313e156579da71fcf281b0883e36674/pycurl-7.45.6-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:c31b390f1e2cd4525828f1bb78c1f825c0aab5d1588228ed71b22c4784bdb593", size = 3499748 },
+    { url = "https://files.pythonhosted.org/packages/22/b0/c5d90c97741e9896c76490edead6056adb87d7083e3bf4761218072cf10d/pycurl-7.45.6-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:942b352b69184cb26920db48e0c5cb95af39874b57dbe27318e60f1e68564e37", size = 3640625 },
+    { url = "https://files.pythonhosted.org/packages/09/63/c7ec2bfedabd71cd71acaed094f3a831ea06198e3b078a199fc353178728/pycurl-7.45.6-cp310-cp310-manylinux_2_28_aarch64.whl", hash = "sha256:3441ee77e830267aa6e2bb43b29fd5f8a6bd6122010c76a6f0bf84462e9ea9c7", size = 4879310 },
+    { url = "https://files.pythonhosted.org/packages/ac/b4/1682fe9d8df5223f6dffa34fe99886ec1a02b136d2cea1f5ec4ad46f1548/pycurl-7.45.6-cp310-cp310-manylinux_2_28_x86_64.whl", hash = "sha256:2a21e13278d7553a04b421676c458449f6c10509bebf04993f35154b06ee2b20", size = 4561745 },
+    { url = "https://files.pythonhosted.org/packages/be/57/0039590b322ef00a1e5664fab888f26bc18b891d0675340d151a106101ca/pycurl-7.45.6-cp310-cp310-win32.whl", hash = "sha256:d0b5501d527901369aba307354530050f56cd102410f2a3bacd192dc12c645e3", size = 2607483 },
+    { url = "https://files.pythonhosted.org/packages/23/c1/7e6c29a814de032d21fe40e679855a6b359f3b2099ab48f9bef346d88b41/pycurl-7.45.6-cp310-cp310-win_amd64.whl", hash = "sha256:abe1b204a2f96f2eebeaf93411f03505b46d151ef6d9d89326e6dece7b3a008a", size = 3135985 },
+    { url = "https://files.pythonhosted.org/packages/85/45/0cc7060dd0a69dbe4806c71460e7a388e2b19097298eb2b4ac4acddea2a1/pycurl-7.45.6-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:6f57ad26d6ab390391ad5030790e3f1a831c1ee54ad3bf969eb378f5957eeb0a", size = 3498594 },
+    { url = "https://files.pythonhosted.org/packages/f4/a9/07a54c81b3fc6d5634e7298f657f9b8fd3d907b1245ef0d16f3b60f67835/pycurl-7.45.6-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:c6fd295f03c928da33a00f56c91765195155d2ac6f12878f6e467830b5dce5f5", size = 3640087 },
+    { url = "https://files.pythonhosted.org/packages/59/bc/fb422d2f7568e5ebdbbf34ecfb3e80303d4fa4679319f7d068fb769c521e/pycurl-7.45.6-cp311-cp311-manylinux_2_28_aarch64.whl", hash = "sha256:334721ce1ccd71ff8e405470768b3d221b4393570ccc493fcbdbef4cd62e91ed", size = 4884164 },
+    { url = "https://files.pythonhosted.org/packages/9e/5a/d6708fbf2727a256983513828250c097aff6a99fc222071a030f108f41ba/pycurl-7.45.6-cp311-cp311-manylinux_2_28_x86_64.whl", hash = "sha256:0cd6b7794268c17f3c660162ed6381769ce0ad260331ef49191418dfc3a2d61a", size = 4566286 },
+    { url = "https://files.pythonhosted.org/packages/e2/c6/ab4e44ac30853b53239af03eeb4e798851c4bfbe182f3aeec29dc7f11e41/pycurl-7.45.6-cp311-cp311-win32.whl", hash = "sha256:357ea634395310085b9d5116226ac5ec218a6ceebf367c2451ebc8d63a6e9939", size = 2607279 },
+    { url = "https://files.pythonhosted.org/packages/c4/90/b26b72fc8e10cd3ab9fb2cfc995106348e2916779de7feb629772f67d8d2/pycurl-7.45.6-cp311-cp311-win_amd64.whl", hash = "sha256:878ae64484db18f8f10ba99bffc83fefb4fe8f5686448754f93ec32fa4e4ee93", size = 3135613 },
+    { url = "https://files.pythonhosted.org/packages/0a/31/01e3ae56940c9aff2886f06a660641ec779d7265aef8b3bd981add67f3a9/pycurl-7.45.6-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:c872d4074360964697c39c1544fe8c91bfecbff27c1cdda1fee5498e5fdadcda", size = 3499118 },
+    { url = "https://files.pythonhosted.org/packages/02/b2/365423ac0f8d13afffb28e6f3d6bf1aae242934f16ae3eda5f3615c0e978/pycurl-7.45.6-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:56d1197eadd5774582b259cde4364357da71542758d8e917f91cc6ed7ed5b262", size = 3640501 },
+    { url = "https://files.pythonhosted.org/packages/67/d6/7d24b60ee878167e3cef6d37fad7d15cf660c0d510539d1417fd59665d1b/pycurl-7.45.6-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:8a99e56d2575aa74c48c0cd08852a65d5fc952798f76a34236256d5589bf5aa0", size = 4893499 },
+    { url = "https://files.pythonhosted.org/packages/24/8c/c84df085310f2489e293adc7880f82339f50dcd6cf9dba2750b81eaf11d4/pycurl-7.45.6-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:c04230b9e9cfdca9cf3eb09a0bec6cf2f084640f1f1ca1929cca51411af85de2", size = 4576761 },
+    { url = "https://files.pythonhosted.org/packages/bc/a2/682c909d93c27b6c6cd369c5427fee87c525e02c56e1a985c150a2d451fd/pycurl-7.45.6-cp312-cp312-win32.whl", hash = "sha256:ae893144b82d72d95c932ebdeb81fc7e9fde758e5ecd5dd10ad5b67f34a8b8ee", size = 2605115 },
+    { url = "https://files.pythonhosted.org/packages/71/42/b1189c859cec1bdb00c93177e8bdee7c75b23a0e30eba41da637b40444c1/pycurl-7.45.6-cp312-cp312-win_amd64.whl", hash = "sha256:56f841b6f2f7a8b2d3051b9ceebd478599dbea3c8d1de8fb9333c895d0c1eea5", size = 3132702 },
+]
+
 [[package]]
 name = "pydantic"
 version = "2.10.6"