test new scheduler

.gitignore

@@ -23,3 +23,9 @@ page_txt.txt
 transcribed_text.txt
 helper/examples/.cache_images/
 helper/examples/images/localtest/
+.env
+TODO.md
+.cache_images/
+traffic_data.db
+ip_data.csv
+data/

app.py

@@ -1,4 +1,14 @@
+import hashlib
+import os
+import shutil
+import sqlite3
+from datetime import datetime
+
 import gradio as gr
+import huggingface_hub
+import pandas as pd
+import pytz
+from apscheduler.schedulers.background import BackgroundScheduler
 
 from helper.gradio_config import css, theme
 from helper.text.text_about import TextAbout
@@ -8,8 +18,78 @@ from helper.text.text_roadmap import TextRoadmap
 from tabs.htr_tool import htr_tool_tab
 from tabs.stepwise_htr_tool import stepwise_htr_tool_tab
 
+DB_FILE = "./traffic_data.db"
+
+TOKEN = os.environ.get("HUB_TOKEN")
+repo = huggingface_hub.Repository(
+    local_dir="data", repo_type="dataset", clone_from="Riksarkivet/traffic_demo_data", use_auth_token=TOKEN
+)
+repo.git_pull()
+
+# Set db to latest
+shutil.copyfile("./data/traffic_data.db", DB_FILE)
+
+
+def hash_ip(ip_address):
+    return hashlib.sha256(ip_address.encode()).hexdigest()
+
+
+# Create table if it doesn't already exist
+db = sqlite3.connect(DB_FILE)
+try:
+    db.execute("SELECT * FROM ip_data").fetchall()
+    db.close()
+except sqlite3.OperationalError:
+    db.execute(
+        """
+        CREATE TABLE ip_data (id INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL,
+                              current_time TIMESTAMP DEFAULT CURRENT_TIMESTAMP NOT NULL,
+                              hashed_ip TEXT)
+        """
+    )
+    db.commit()
+    db.close()
+
+
+def current_time_sw():
+    swedish_tz = pytz.timezone("Europe/Stockholm")
+    return datetime.now(swedish_tz).strftime("%Y-%m-%d %H:%M:%S")
+
+
+def add_ip_data(request: gr.Request):
+    host = request.client.host
+    hashed_ip = hash_ip(host)
+
+    db = sqlite3.connect(DB_FILE)
+    cursor = db.cursor()
+    cursor.execute("INSERT INTO ip_data(current_time, hashed_ip) VALUES(?,?)", [current_time_sw(), hashed_ip])
+    db.commit()
+    db.close()
+
+
+def backup_db():
+    shutil.copyfile(DB_FILE, "./data/traffic_data.db")
+    db = sqlite3.connect(DB_FILE)
+    ip_data = db.execute("SELECT * FROM ip_data").fetchall()
+    pd.DataFrame(ip_data, columns=["id", "current_time", "hashed_ip"]).to_csv("./data/ip_data.csv", index=False)
+
+    print("updating traffic_data")
+    repo.push_to_hub(blocking=False, commit_message=f"Updating data at {datetime.now()}")
+
+
+scheduler = BackgroundScheduler()
+scheduler.add_job(func=backup_db, trigger="interval", seconds=60)
+scheduler.start()
+
+
 with gr.Blocks(title="HTR Riksarkivet", theme=theme, css=css) as demo:
-    gr.Markdown(TextApp.title_markdown)
+    with gr.Row():
+        with gr.Column(scale=1):
+            text_ip_output = gr.Markdown()
+        with gr.Column(scale=1):
+            gr.Markdown(TextApp.title_markdown)
+        with gr.Column(scale=1):
+            gr.Markdown(TextApp.title_markdown_img)
 
     with gr.Tabs():
         with gr.Tab("HTR Tool"):
@@ -18,61 +98,16 @@ with gr.Blocks(title="HTR Riksarkivet", theme=theme, css=css) as demo:
         with gr.Tab("Stepwise HTR Tool"):
             stepwise_htr_tool_tab.render()
 
-        with gr.Tab("How to use"):
-            with gr.Tabs():
-                with gr.Tab("HTR Tool"):
-                    with gr.Row(equal_height=False):
-                        with gr.Column():
-                            gr.Markdown(TextHowTo.htr_tool)
-                        with gr.Column():
-                            gr.Markdown(TextHowTo.both_htr_tool_video)
-                            gr.Video(
-                                value="https://github.com/Borg93/htr_gradio_file_placeholder/raw/main/htr_tool_media_cut.mp4",
-                                label="How to use HTR Tool",
-                            )
-                            gr.Markdown(TextHowTo.reach_out)
-
-                with gr.Tab("Stepwise HTR Tool"):
-                    with gr.Row(equal_height=False):
-                        gr.Markdown(TextHowTo.stepwise_htr_tool)
-                    with gr.Row():
-                        gr.Markdown(TextHowTo.stepwise_htr_tool_tab_intro)
-                    with gr.Row():
-                        with gr.Accordion("The tabs for the Stepwise HTR Tool:", open=True):
-                            with gr.Tabs():
-                                with gr.Tab("1. Region Segmentation"):
-                                    gr.Markdown(TextHowTo.stepwise_htr_tool_tab1)
-                                with gr.Tab("2. Line Segmentation"):
-                                    gr.Markdown(TextHowTo.stepwise_htr_tool_tab2)
-                                with gr.Tab("3. Transcribe Text"):
-                                    gr.Markdown(TextHowTo.stepwise_htr_tool_tab3)
-                                with gr.Tab("4. Explore Results"):
-                                    gr.Markdown(TextHowTo.stepwise_htr_tool_tab4)
-                    gr.Markdown(TextHowTo.stepwise_htr_tool_end)
-
-                with gr.Tab("API & Duplicate for Privat use"):
-                    with gr.Row():
-                        with gr.Column():
-                            gr.Markdown(TextHowTo.htr_tool_api_text)
-                            gr.Code(
-                                value=TextHowTo.code_for_api,
-                                language="python",
-                                interactive=False,
-                                show_label=False,
-                            )
-                        with gr.Column():
-                            gr.Markdown(TextHowTo.duplicatin_space_htr_text)
-                            gr.Markdown(TextHowTo.figure_htr_hardware)
-                            gr.Markdown(TextHowTo.duplicatin_for_privat)
-
         with gr.Tab("About"):
             with gr.Tabs():
                 with gr.Tab("Project"):
                     with gr.Row():
                         with gr.Column():
-                            gr.Markdown(TextAbout.intro_and_pipeline_overview_text)
+                            gr.Markdown(TextAbout.intro_text)
                         with gr.Column():
                             gr.Markdown(TextAbout.text_src_code_data_models)
+                    with gr.Row():
+                        gr.Markdown(TextAbout.pipeline_overview_text)
                     with gr.Row():
                         with gr.Tabs():
                             with gr.Tab("I. Binarization"):
@@ -88,6 +123,21 @@ with gr.Blocks(title="HTR Riksarkivet", theme=theme, css=css) as demo:
                     with gr.Row():
                         gr.Markdown(TextRoadmap.text_contribution)
 
+                with gr.Tab("API & Duplicate for Privat use"):
+                    with gr.Row():
+                        with gr.Column():
+                            gr.Markdown(TextHowTo.htr_tool_api_text)
+                            gr.Code(
+                                value=TextHowTo.code_for_api,
+                                language="python",
+                                interactive=False,
+                                show_label=False,
+                            )
+                        with gr.Column():
+                            gr.Markdown(TextHowTo.duplicatin_space_htr_text)
+                            gr.Markdown(TextHowTo.figure_htr_hardware)
+                            gr.Markdown(TextHowTo.duplicatin_for_privat)
+
                 with gr.Tab("Roadmap"):
                     with gr.Row():
                         with gr.Column():
@@ -95,7 +145,7 @@ with gr.Blocks(title="HTR Riksarkivet", theme=theme, css=css) as demo:
                         with gr.Column():
                             gr.Markdown(TextRoadmap.discussion)
 
-    # demo.load(None, None, None, _js=js)
+    demo.load(add_ip_data)
 
 
 demo.queue(concurrency_count=2, max_size=2)

helper/text/text_about.py

@@ -1,6 +1,6 @@
 class TextAbout:
     # About text
-    intro_and_pipeline_overview_text = """
+    intro_text = """
 
     ## Introduction
     The Swedish National Archives introduces a demonstrational end-to-end HTR (Handwritten Text Recognition) pipeline. This pipeline comprises two instance segmentation models: one designated for segmenting text-regions and another for isolating text-lines within these regions, coupled with an HTR model for image-to-text transcription. The objective of this project is to establish a generic pipeline capable of processing running-text documents spanning from 1600 to 1900.
@@ -12,10 +12,14 @@ class TextAbout:
     -  Navigate to the > **About** > **Roadmap**. 
 
     To understand how to utilize this application through a REST API, self-host or via Docker, 
-    - Navigate to the > **How to Use** > **API & Duplicate for Private Use**.
+    - Navigate to the > **About** > **How to Use** > **API & Duplicate for Private Use**.
+    
+    """
 
     ## The Pipeline in Overview
-
+    pipeline_overview_text = """
+    ## The Pipeline in Overview
+    
     The steps in the pipeline can be seen below as follows:
     """
 

helper/text/text_app.py

@@ -1,12 +1,16 @@
 class TextApp:
     title_markdown = """
 
-    <img src="https://raw.githubusercontent.com/Borg93/Riksarkivet_docs/main/docs/assets/fav-removebg-preview.png" width="4%" align="right" margin-right="100" />
 
-    <h1><center> Handwritten Text Recognition Tool </center></h1>
+    <h1><center> HTRflow - Demo </center></h1>
 
     <h3><center> Swedish National Archives - Riksarkivet </center></h3>"""
 
+    title_markdown_img = """
+    <img src="https://raw.githubusercontent.com/Borg93/Riksarkivet_docs/main/docs/assets/fav-removebg-preview.png" width="13%" align="right" margin-right="100" />
+
+    """
+
 
 if __name__ == "__main__":
     pass

requirements.txt

@@ -7,9 +7,13 @@ numpy==1.25.0
 opencv-python-headless==4.7.0.72
 Jinja2==3.1.2
 transformers==4.30.2
-huggingface-hub==0.15.1
+huggingface-hub
 datasets==2.14.5
 requests==2.31.0
+apscheduler
+pytz
+jiwer
+evaluate
 # pillow==9.5.0
 
 

src/htr_pipeline/gradio_backend.py

@@ -117,7 +117,7 @@ class CustomTrack:
             gr.update(visible=True),
         )
 
-    def transcribe_text(self, df, images):
+    def transcribe_text(self, images):
         gr.Info("Running Transcribe Lines")
         transcription_temp_list_with_score = []
         mapping_dict = {}
@@ -142,11 +142,11 @@ class CustomTrack:
                 transcription_temp_list_with_score, columns=["Transcribed text", "Pred score"]
             )
 
+            joined_transcription_temp_list = "\n".join([tup[0] for tup in transcription_temp_list_with_score])
+
             mapping_dict[transcribed_text] = image
 
-            yield df_trans_explore[
-                ["Transcribed text"]
-            ], df_trans_explore, mapping_dict, bool_to_show_control_results_transcribe, bool_to_show_placeholder
+            yield joined_transcription_temp_list, df_trans_explore, mapping_dict, bool_to_show_control_results_transcribe, bool_to_show_placeholder
 
     def get_select_index_image(self, images_from_gallery, evt: gr.SelectData):
         return images_from_gallery[evt.index]["name"]

tabs/htr_tool.py

@@ -155,6 +155,15 @@ with gr.Blocks() as htr_tool_tab:
         api_name="predict",
     )
 
+    def dummy_update_htr_tool_transcriber_model_dropdown(htr_tool_transcriber_model_dropdown):
+        return gr.update(value="Riksarkivet/satrn_htr")
+
+    htr_tool_transcriber_model_dropdown.change(
+        fn=dummy_update_htr_tool_transcriber_model_dropdown,
+        inputs=htr_tool_transcriber_model_dropdown,
+        outputs=htr_tool_transcriber_model_dropdown,
+    )
+
     def update_selected_tab_output_and_setting():
         return gr.update(visible=True), gr.update(visible=False)
 

tabs/stepwise_htr_tool.py

@@ -1,9 +1,11 @@
 import os
 import shutil
 
+import evaluate
 import gradio as gr
 
 from helper.examples.examples import DemoImages
+from helper.text.text_howto import TextHowTo
 from src.htr_pipeline.gradio_backend import CustomTrack, SingletonModelLoader
 
 model_loader = SingletonModelLoader()
@@ -12,9 +14,29 @@ custom_track = CustomTrack(model_loader)
 
 images_for_demo = DemoImages()
 
+cer_metric = evaluate.load("cer")
+
+
 with gr.Blocks() as stepwise_htr_tool_tab:
     with gr.Tabs():
         with gr.Tab("1. Region Segmentation"):
+            with gr.Row():
+                with gr.Accordion("Info", open=False) as example_accord:
+                    with gr.Row(equal_height=False):
+                        gr.Markdown(TextHowTo.stepwise_htr_tool)
+                    with gr.Row():
+                        gr.Markdown(TextHowTo.stepwise_htr_tool_tab_intro)
+                    with gr.Row():
+                        with gr.Tabs():
+                            with gr.Tab("1. Region Segmentation"):
+                                gr.Markdown(TextHowTo.stepwise_htr_tool_tab1)
+                            with gr.Tab("2. Line Segmentation"):
+                                gr.Markdown(TextHowTo.stepwise_htr_tool_tab2)
+                            with gr.Tab("3. Transcribe Text"):
+                                gr.Markdown(TextHowTo.stepwise_htr_tool_tab3)
+                            with gr.Tab("4. Explore Results"):
+                                gr.Markdown(TextHowTo.stepwise_htr_tool_tab4)
+                    gr.Markdown(TextHowTo.stepwise_htr_tool_end)
             with gr.Row():
                 with gr.Column(scale=2):
                     vis_data_folder_placeholder = gr.Markdown(visible=False)
@@ -25,38 +47,9 @@ with gr.Blocks() as stepwise_htr_tool_tab:
                             label="Image to Region segment",
                             # type="numpy",
                             tool="editor",
-                            height=350,
+                            height=400,
                         )
 
-                    with gr.Accordion("Region segment settings:", open=False):
-                        with gr.Row():
-                            reg_pred_score_threshold_slider = gr.Slider(
-                                minimum=0.4,
-                                maximum=1,
-                                value=0.5,
-                                step=0.05,
-                                label="P-threshold",
-                                info="""Filter and determine the confidence score 
-                                                required for a prediction score to be considered""",
-                            )
-                            reg_containments_threshold_slider = gr.Slider(
-                                minimum=0,
-                                maximum=1,
-                                value=0.5,
-                                step=0.05,
-                                label="C-threshold",
-                                info="""The minimum required overlap or similarity 
-                                                for a detected region or object to be considered valid""",
-                            )
-
-                        with gr.Row():
-                            region_segment_model_dropdown = gr.Dropdown(
-                                choices=["Riksarkivet/RmtDet_region"],
-                                value="Riksarkivet/RmtDet_region",
-                                label="Region segment model",
-                                info="Will add more models later!",
-                            )
-
                     with gr.Row():
                         clear_button = gr.Button("Clear", variant="secondary", elem_id="clear_button")
 
@@ -66,7 +59,36 @@ with gr.Blocks() as stepwise_htr_tool_tab:
                             elem_id="region_segment_button",
                         )
 
-                    with gr.Row():
+                    with gr.Group():
+                        with gr.Accordion("Region segment settings:", open=False):
+                            with gr.Row():
+                                reg_pred_score_threshold_slider = gr.Slider(
+                                    minimum=0.4,
+                                    maximum=1,
+                                    value=0.5,
+                                    step=0.05,
+                                    label="P-threshold",
+                                    info="""Filter and determine the confidence score 
+                                                    required for a prediction score to be considered""",
+                                )
+                                reg_containments_threshold_slider = gr.Slider(
+                                    minimum=0,
+                                    maximum=1,
+                                    value=0.5,
+                                    step=0.05,
+                                    label="C-threshold",
+                                    info="""The minimum required overlap or similarity 
+                                                    for a detected region or object to be considered valid""",
+                                )
+
+                            with gr.Row():
+                                region_segment_model_dropdown = gr.Dropdown(
+                                    choices=["Riksarkivet/RmtDet_region"],
+                                    value="Riksarkivet/RmtDet_region",
+                                    label="Region segment model",
+                                    info="Will add more models later!",
+                                )
+
                         with gr.Accordion("Example images to use:", open=False) as example_accord:
                             gr.Examples(
                                 examples=images_for_demo.examples_list,
@@ -76,7 +98,7 @@ with gr.Blocks() as stepwise_htr_tool_tab:
                             )
 
                 with gr.Column(scale=3):
-                    output_region_image = gr.Image(label="Segmented regions", type="numpy", height=600)
+                    output_region_image = gr.Image(label="Segmented regions", type="numpy", height=550)
 
         ##############################################
         with gr.Tab("2. Line Segmentation"):
@@ -188,14 +210,11 @@ with gr.Blocks() as stepwise_htr_tool_tab:
 
                 with gr.Column(scale=3):
                     with gr.Row():
-                        transcribed_text_df = gr.Dataframe(
-                            headers=["Transcribed text"],
-                            max_rows=14,
-                            col_count=(1, "fixed"),
-                            wrap=True,
-                            interactive=False,
-                            overflow_row_behaviour="paginate",
-                            height=600,
+                        transcribed_text = gr.Textbox(
+                            label="Transcribed text",
+                            info="Transcribed text is being streamed back from the HTR-model",
+                            lines=25,
+                            value="",
                         )
 
         #####################################
@@ -219,18 +238,27 @@ with gr.Blocks() as stepwise_htr_tool_tab:
                             columns=[3],
                             rows=[3],
                             # object_fit="contain",
-                            height=300,
+                            height=250,
                             preview=True,
                             container=False,
                         )
 
                     dataframe_text_index = gr.Textbox(
                         label="Text from DataFrame selection",
-                        info="Click on a dataframe cell to view the corresponding transcribed text line crop. You can also sort the dataframe to easily locate specific entries.",
-                        lines=2,
+                        placeholder="Select row from the DataFrame.",
                         interactive=False,
                     )
 
+                    gt_text_index = gr.Textbox(
+                        label="Ground Truth",
+                        placeholder="Provide the ground truth, if available.",
+                        interactive=True,
+                    )
+                    with gr.Row(equal_height=False):
+                        calc_cer_button = gr.Button("Calculate CER", variant="primary", visible=True)
+
+                        cer_output = gr.Textbox(label="CER:")
+
                 with gr.Column(scale=1, visible=True):
                     mapping_dict = gr.Variable()
                     transcribed_text_df_finish = gr.Dataframe(
@@ -279,9 +307,9 @@ with gr.Blocks() as stepwise_htr_tool_tab:
 
     transcribe_button.click(
         custom_track.transcribe_text,
-        inputs=[transcribed_text_df, inputs_lines_to_transcribe],
+        inputs=[inputs_lines_to_transcribe],
         outputs=[
-            transcribed_text_df,
+            transcribed_text,
             transcribed_text_df_finish,
             mapping_dict,
             # Hide
@@ -290,6 +318,14 @@ with gr.Blocks() as stepwise_htr_tool_tab:
         ],
     )
 
+    def compute_cer(dataframe_text_index, gt_text_index):
+        if gt_text_index is not None and gt_text_index.strip() != "":
+            return cer_metric.compute(predictions=[dataframe_text_index], references=[gt_text_index])
+        else:
+            return "Ground truth not provided"
+
+    calc_cer_button.click(compute_cer, inputs=[dataframe_text_index, gt_text_index], outputs=cer_output)
+
     clear_button.click(
         lambda: (
             (shutil.rmtree("./vis_data") if os.path.exists("./vis_data") else None, None)[1],
@@ -318,7 +354,7 @@ with gr.Blocks() as stepwise_htr_tool_tab:
             control_line_segment,
             output_line_from_region,
             inputs_lines_to_transcribe,
-            transcribed_text_df,
+            transcribed_text,
             control_htr,
             inputs_lines_to_transcribe,
             image_placeholder_htr,