ui/main_page.py

import gradio as gr
from functools import partial
from gradio_modal import Modal
from data.lang2eng_map import lang2eng_mapping
from data.words_map import words_mapping
import datetime
from pyuca import Collator


def sort_with_pyuca(strings):
    collator = Collator()
    return sorted(strings, key=collator.sort_key)


def build_main_page(concepts_dict, metadata_dict, local_storage):
    try:
        country, lang, _, _ = local_storage.value
        if not country and not lang:
            country, lang = "USA", "English"
    except (TypeError, ValueError):
        country, lang = "USA", "English"
    with gr.Column(visible=False, elem_id="main_page") as main_ui_placeholder:
        # Read the markdown file
        with open(metadata_dict["USA"]["English"]["Task"], "r", encoding="utf-8") as f:
            TASK_TEXT = f.read()

        with open(metadata_dict["USA"]["English"]["Instructions"], "r", encoding="utf-8") as f:
            INST_TEXT = f.read()

        gr.Markdown("# Wonders: Multicultural Images and Captions for Inclusive AI")
        gr.Markdown("Help us collect culturally rich images and captions to improve AI for everyone!")
        gr.Markdown("Choose a concept, upload an image showing it in a typical local situation, and write a detailed description about the image.<br> For more information, see the overviews below or click on 'Show instructions' for detailed guidelines.")
        gr.Markdown(""" \
            ### Notes
            - By participating in this task, you agree to donate images and captions under [CC BY license](https://creativecommons.org/licenses/by/4.0/), which will allow us to share the data with others to improve AI.
            - You will be able to **update or remove** any provided data at any time if you wish so.
            - Participants that provide more data, across more concepts and of high quality (see "Instructions" button) will be **rewarded** (see TODO).
            """
        )
        # with gr.Accordion("📘 Show task overview", open=False):
        intro_text_inp = gr.Markdown(TASK_TEXT, visible=False)

        # gr.Markdown("## Data Collection")
        with gr.Row(equal_height=True):
            country_inp = gr.Textbox(label="Country", elem_id="country_inp", interactive=False)
            language_inp = gr.Textbox(label="Language", elem_id="language_inp", interactive=False)
            username_inp = gr.Textbox(label="email (optional)", type="email", elem_id="username_inp", interactive=False)
            password_inp = gr.Textbox(label="password (optional)", type="password", elem_id="password_inp", interactive=False)
            exit_btn = gr.Button("⬅️ Change Language", elem_id="exit_btn", elem_classes=["compact-btn"])

        with gr.Row(equal_height=True, ):
            with gr.Column():
                # Main category and concept row - SINGLE SELECTION
                with gr.Group():
                    gr.Markdown("### Main Concept")
                    gr.Markdown("""\
                        Select the main concept and category for which you want to share data.<br>
                        If you want to add a new concept, you can just type it in the same box (after selecting its category).
                        """
                    )
                    with gr.Row():
                        # Use concepts_dict to populate the dropdown
                        categories = concepts_dict["USA"]["English"]
                        category_btn = gr.Dropdown(choices=sort_with_pyuca(categories.keys()), interactive=True, label="Main Category", 
                                                allow_custom_value=False, elem_id="category_btn", multiselect=False, value=None)
                        concept_btn = gr.Dropdown(choices=[], interactive=True, label="Main Concept", 
                                                allow_custom_value=True, elem_id="concept_btn", multiselect=False)
                
                with gr.Group():
                    gr.Markdown("### Image")
                    with gr.Accordion("📘 An image of the main concept in typical contexts (click to read more)", open=False):
                        gr.Markdown("""\=
                            An image where the *main concept is clearly visible*
                            - in typical contexts in your country, and
                            - with other objects/people when possible
                            
                            Pictures taken by yourself are preferred.
                            If you want to use an image from the Web, directly add its URL below.
                            - Please use publicly-licensed images (e.g., from pexels.com or freepik.com)
                            
                            For more details, click on “Show instructions” below.
                            """
                        )
                    image_url_inp = gr.Textbox(label="Image URL (Optional, if not uploading an image)", type="text", elem_id="image_url_inp")                
                    image_inp = gr.Image(label="Image", elem_id="image_inp", format="png", height=512, width=768)
                    with gr.Row():
                        hide_all_faces_btn = gr.Button("👤 Hide All Faces", elem_id="hide_all_faces_btn")
                        hide_faces_btn = gr.Button("👤 Hide Specific Faces", elem_id="hide_faces_btn")
                        unhide_faces_btn = gr.Button("👀 Unhide Faces", elem_id="unhide_faces_btn")
                    with gr.Row():
                        rotate_left_btn = gr.Button("↺ Rotate Left", elem_id="rotate_left_btn")
                        rotate_right_btn = gr.Button("↻ Rotate Right", elem_id="rotate_right_btn")
                        reset_image_btn = gr.Button("🔄 Reset Image", elem_id="reset_image_btn")

            with gr.Column():
                # short_caption_inp = gr.Textbox(lines=2, label="Short Description", elem_id="short_caption_inp")
                with gr.Group():
                    gr.Markdown("### Description")
                    with gr.Accordion("📘 A detailed description (around 40 words covering who?, what? and where?) for the image (click to read more)", open=False):
                        gr.Markdown("""\
                            A long image description (around 40 words) with extensive and detailed visual information.<br>
                            Descriptions must be *objective*: focus on how you would describe the image to someone who can’t see it, without your own opinions/speculations.
                            The text needs to *include the main concept* and describes the content of the image in detail by including:
                            - **Who?** The visual *appearance* and observable *emotions* (e.g., “is smiling”) of persons and animals.
                            - **What?** The *actions* performed in the image.
                            - **Where?** The *setting* of the image, including the size, color and relationships between objects.
                            
                            For more details, click on “Show instructions” below.
                            """
                        )
                    long_caption_inp = gr.Textbox(lines=6, label="Description", elem_id="long_caption_inp")
                    num_words_inp = gr.Textbox(lines=1, label="Number of words", elem_id="num_words", interactive=False, value=0)
                    # num_words_inp = gr.Markdown("Number of words", elem_id="num_words")
                #########################################################
                with Modal(visible=False, allow_user_close=False) as modal_vlm:

                    question = gr.Markdown("Would you like to see if a VLM can generate a culturally aware description for your uploaded concept?")
                    with gr.Row():
                        gen_button = gr.Button("Yes", variant="primary", elem_id="generate_answer_btn")
                        vlm_no_btn = gr.Button("No")
                        vlm_cancel_btn = gr.Button("Cancel")
                    # ["SmolVLM-500M", "Qwen2.5-VL-7B", "InternVL3_5-8B", "Gemma3-4B"]
                    vlm_model_dropdown = gr.Dropdown(
                        ["SmolVLM-500M"], value="SmolVLM-500M", multiselect=False, label="VLM Model", info="Select the VLM model to use for generating the description."
                    )
                    vlm_output = gr.Textbox(lines=6, label="Generated description", elem_id="vlm_output", interactive=False)
                    vlm_feedback = gr.Radio(["Yes 👍", "No 👎"], label="Do you think the generated description is accurate within the cultural context of your country?", visible=False, elem_id="vlm_feedback", interactive=True)
                    vlm_done_btn = gr.Button("Complete Submission", visible=False)

                with Modal(visible=False, allow_user_close=False) as modal_submit:

                    gr.Markdown("⚠️ You've already generated a caption for this image. An optional description with the VLM can only be generated once. Would you like to proceed and submit your modified data?")
                    with gr.Row():
                        submit_yes = gr.Button("Yes", variant="primary", elem_id="submit_confirm_yes")
                        submit_no = gr.Button("No", variant="stop", elem_id="submit_confirm_no")

                    # with gr.Group():
                    #     gr.Markdown("### VLM Generation (Optional)")
                    #     with gr.Accordion("📘 Click here if you want to get a generated answer from a small vlm", open=False):
                    #         gen_button = gr.Button("Generate Answer", variant="primary", elem_id="generate_answer_btn")
                    #         vlm_output = gr.Textbox(lines=6, label="Generated Answer", elem_id="vlm_output", interactive=False)
                    #         vlm_feedback = gr.Radio(["Yes 👍", "No 👎"], label="Do you like the generated caption?", visible=False, elem_id="vlm_feedback", interactive=True)
                ##########################################################
                
                categories_list = sort_with_pyuca(list(concepts_dict["USA"]["English"].keys()))
                
                def create_category_dropdown(category, index):
                    original_category = category
                    
                    if lang in words_mapping:
                        display_category = words_mapping[lang].get(original_category, original_category)
                    else:
                        display_category = original_category
                    
                    category_choices = concepts_dict[country][lang2eng_mapping.get(lang, lang)][original_category]
                    sorted_choices = sort_with_pyuca(category_choices)
                    
                    dropdown = gr.Dropdown(
                        choices=sorted_choices, 
                        interactive=True, 
                        label=display_category,
                        allow_custom_value=True, 
                        elem_id=f"category_{index+1}_concepts_btn", 
                        multiselect=True,
                        value=None
                    )
                    return dropdown
                
                category_concept_dropdowns = []
                
                with gr.Group():
                    gr.Markdown("### Additional concepts (optional)")
                    gr.Markdown("Tag any other concept that is visible in the image.")
                    # First row - categories 1 and 2
                    with gr.Row():
                        dropdown1 = create_category_dropdown(categories_list[0], 0)
                        category_concept_dropdowns.append(dropdown1)
                        
                        dropdown2 = create_category_dropdown(categories_list[1], 1)
                        category_concept_dropdowns.append(dropdown2)

                    # Second row - categories 3 and 4
                    # with gr.Row():
                        dropdown3 = create_category_dropdown(categories_list[2], 2)
                        category_concept_dropdowns.append(dropdown3)
                        
                        dropdown4 = create_category_dropdown(categories_list[3], 3)
                        category_concept_dropdowns.append(dropdown4)

                        dropdown5 = create_category_dropdown(categories_list[4], 4)
                        category_concept_dropdowns.append(dropdown5)

                # Third row - category 5 and instructions button
                with gr.Row(equal_height=True):
                    # dropdown5 = create_category_dropdown(categories_list[4], 4)
                    # category_concept_dropdowns.append(dropdown5)
                    
                    instruct_btn = gr.Button("📘 Show Instructions")
                    with Modal(visible=False) as modal:
                        intro_text_inst_inp = gr.Markdown(INST_TEXT)              

                with gr.Column():
                    # with gr.Row():
                    #     instruct_btn = gr.Button("📘 Show Instructions")
                    #     with Modal(visible=False) as modal:
                    #         intro_text_inst_inp = gr.Markdown(INST_TEXT)
                    with gr.Row(equal_height=True):
                        clear_btn = gr.Button("Clear", variant="huggingface", elem_id="clear_btn")
                        with Modal(visible=False, allow_user_close=False) as modal_saving:
                            modal_saving_text = gr.Markdown("⏳ Please wait while your submission is being saved.")
                        with Modal(visible=False) as modal_data_saved:
                            modal_data_saved_text = gr.Markdown("Your data has been saved successfully. The data in the table below will be updated shortly. You can now close this window.")
                        submit_btn = gr.Button("Submit", variant="primary", interactive=False, elem_id="submit_btn")
                        with Modal(visible=False) as modal_exclude_confirm:
                            gr.Markdown("## Are you sure you want to exclude this example?")
                            gr.Markdown("This action will permanently delete the example.")
                            with gr.Row():
                                cancel_exclude_btn = gr.Button("Cancel")
                                confirm_exclude_btn = gr.Button("Yes, delete", variant="stop")
                        exclude_btn = gr.Button("Exclude Selected Example", variant="stop", visible=True)
                with gr.Column():
                    timestamp_btn = gr.Textbox(datetime.datetime.now(), label="Timestamp", visible=False, elem_id="timestamp_btn", interactive=False) # FIXME visible=False)
                    exampleid_btn = gr.Textbox(label="ID", visible=False, elem_id="example_id", interactive=False) # FIXME visible=False)

    output_dict = {
        "main_ui_placeholder": main_ui_placeholder,
        "country_inp": country_inp,
        "language_inp": language_inp,
        "username_inp": username_inp,
        "password_inp": password_inp,
        "image_inp": image_inp,
        "image_url_inp": image_url_inp,
        "long_caption_inp": long_caption_inp,
        "num_words_inp": num_words_inp,
        "category_btn": category_btn,
        "concept_btn": concept_btn,
        "category_concept_dropdowns": category_concept_dropdowns,
        "category_1_concepts": category_concept_dropdowns[0],
        "category_2_concepts": category_concept_dropdowns[1],
        "category_3_concepts": category_concept_dropdowns[2],
        "category_4_concepts": category_concept_dropdowns[3],
        "category_5_concepts": category_concept_dropdowns[4],   
        "instruct_btn": instruct_btn,
        "clear_btn": clear_btn,
        "submit_btn": submit_btn,
        "modal": modal,
        "modal_saving": modal_saving,
        "modal_data_saved": modal_data_saved,
        "timestamp_btn": timestamp_btn,
        "exampleid_btn": exampleid_btn,
        "exit_btn": exit_btn,
        "intro_text_inp": intro_text_inp,
        "intro_text_inst_inp": intro_text_inst_inp,
        "modal_saving_text": modal_saving_text,
        "modal_data_saved_text": modal_data_saved_text,
        "hide_faces_btn": hide_faces_btn,
        "hide_all_faces_btn": hide_all_faces_btn,
        "unhide_faces_btn": unhide_faces_btn,
        "rotate_left_btn": rotate_left_btn,
        "rotate_right_btn": rotate_right_btn,
        "reset_image_btn": reset_image_btn,
        "exclude_btn": exclude_btn,
        "modal_exclude_confirm": modal_exclude_confirm,
        "cancel_exclude_btn": cancel_exclude_btn,
        "confirm_exclude_btn": confirm_exclude_btn,
        "vlm_output": vlm_output,
        "gen_button": gen_button,
        "vlm_feedback": vlm_feedback,
        "modal_vlm": modal_vlm,
        "vlm_no_btn": vlm_no_btn,
        "vlm_done_btn": vlm_done_btn,
        "submit_yes": submit_yes,
        "submit_no": submit_no,
        "modal_submit": modal_submit,
        "vlm_cancel_btn": vlm_cancel_btn,
        "vlm_model_dropdown": vlm_model_dropdown
    }
    return output_dict