app.py

import gradio as gr

def process_submission(input_img, text_answer, multiple_choice, city, country, se_asia_relevance, culture_knowledge, native_caption, english_caption):
    # Just return the inputs to demonstrate they were received
    # No ML pipeline processing here
    return input_img, f"Your text response: {text_answer}", f"Your selected option: {multiple_choice}", f"Location: {city}, {country}", f"SE Asia relevance: {se_asia_relevance}", f"Cultural knowledge source: {culture_knowledge}", f"Native caption: {native_caption}", f"English caption: {english_caption}"

gradio_app = gr.Interface(
    process_submission,
    inputs=[
        gr.Image(label="Upload an image", sources=['upload', 'webcam'], type="pil"),
        gr.Textbox(label="The image portrays culturally-relevant information in:", placeholder="what culture does this image represent?"),
        # gr.Radio(choices=["North America", "South America", "Europe", "Africa", "Asia", "Australia"], 
        #          label="Which continent is most represented in this image?"),
        gr.Textbox(label="City where the image was taken:", placeholder="Enter city name"),
        gr.Textbox(label="Country where the image was taken:", placeholder="Enter country name"),
        gr.Radio(
            choices=[
                "Yes. Unique to South East",
                "Yes, people will likely think of South East when seeing the picture, but it may have low degree of similarity to other cultures.",
                "Maybe, this culture did not originate from South East, but it's quite dominant in South East",
                "Not really. It has some affiliation to South East, but actually does not represent South East or has stronger affiliation to cultures outside South East",
                "No. Totally unrelated to South East"
            ],
            label="Is the image culturally relevant in South-East Asia?"
        ),
        gr.Radio(
            choices=[
                "I'm from this country/culture",
                "I checked online resources (e.g., Wikipedia, articles, blogs)"
            ],
            label="How do you know about this culture?",
            info="Please do not consult LLMs (e.g., GPT-4o, Claude, Command-R, etc.)"
        ),
        gr.Textbox(label="Caption in Native Language:", placeholder="Enter caption in the native language of the culture depicted"),
        gr.Textbox(label="English Caption:", placeholder="Enter caption in English")
    ],
    outputs=[
        gr.Image(label="Submitted Image"),
        gr.Text(label="Text Response"),
        gr.Text(label="Multiple Choice Response"),
        gr.Text(label="Location Information"),
        gr.Text(label="South-East Asia Cultural Relevance"),
        gr.Text(label="Cultural Knowledge Source"),
        gr.Text(label="Native Language Caption"),
        gr.Text(label="English Caption")
    ],
    title="Image Cultural Analysis",
    description="Upload an image and answer questions about its cultural significance."
)

if __name__ == "__main__":
    gradio_app.launch()