app.py

import pandas as pd
import gradio as gr


data = {
    "Method": [
        "AIN-7B", "GPT-4o", "GPT-4o-mini", "Qwen2-VL-7B", "Gemini-1.5-Pro", "Gemini-1.5-Flash", 
        "LLaVa-OneVision-7B", "Pangea-7B-Instruct", "Qwen2-VL-2B", "InternVL2-8B", "LLaVa-NeXt-7B", "Maya-8B"
    ],
    "MM Understanding & Reasoning": [
        56.78, 55.15, 48.83, 48.76, 46.67, 45.58, 42.90, 40.09, 40.59, 30.41, 26.33, 39.07
    ],
    "OCR & Document Understanding": [
        72.35, 54.98, 39.38, 42.73, 36.59, 33.59, 31.35, 17.75, 25.68, 15.91, 19.12, 26.70
    ],
    "Video Understanding": [
        64.09, 69.65, 66.28, 61.97, 42.94, 53.31, 29.41, 49.01, 38.90, 51.42, 44.90, 47.23
    ],
    "Remote Sensing Understanding": [
        45.92, 27.36, 16.93, 21.30, 17.07, 14.95, 10.72, 6.67, 12.56, 5.36, 8.33, 27.53
    ],
    "Charts & Diagram Understanding": [
        64.10, 62.35, 56.37, 54.67, 47.06, 48.25, 40.86, 38.75, 27.83, 30.27, 27.56, 34.25
    ],
    "Agro Specific": [
        85.05, 80.75, 78.80, 79.32, 72.12, 76.06, 75.03, 74.51, 52.02, 44.47, 42.00, 70.61
    ],
    "Cultural Specific Understanding": [
        78.09, 80.86, 65.92, 75.96, 56.24, 46.54, 66.02, 20.34, 34.27, 20.88, 28.30, 57.42
    ],
    "Medical Imaging": [
        43.77, 49.91, 47.37, 35.81, 33.77, 42.86, 27.29, 31.99, 29.12, 29.48, 22.54, 31.57
    ],
    
}


# data = {
#     "Method": [
#         "GPT-4o", "GPT-4o-mini", "Qwen2-VL-7B", "Gemini-1.5-Pro", "Gemini-1.5-Flash", 
#         "LLaVa-OneVision-7B", "Pangea-7B-Instruct", "Qwen2-VL-2B", "InternVL2-8B", "LLaVa-NeXt-7B", "Maya-8B"
#     ],
#     "MM Understanding & Reasoning": [
#         57.90, 48.82, 51.35, 46.67, 45.58, 42.90, 40.09, 40.59, 30.41, 26.33, 39.07
#     ],
#     "OCR & Document Understanding": [
#         59.11, 42.89, 49.06, 36.59, 33.59, 31.35, 17.75, 25.68, 15.91, 19.12, 26.70
#     ],
#     "Charts & Diagram Understanding": [
#         73.57, 64.98, 55.39, 47.06, 48.25, 40.86, 38.75, 27.83, 30.27, 27.56, 34.25
#     ],
#     "Video Understanding": [
#         74.27, 68.11, 62.64, 42.94, 53.31, 29.41, 49.01, 38.90, 51.42, 44.90, 47.23
#     ],
#     "Cultural Specific Understanding": [
#         80.86, 65.92, 75.64, 56.24, 46.54, 66.02, 20.34, 34.27, 20.88, 28.30, 57.42
#     ],
#     "Medical Imaging": [
#         49.90, 47.37, 39.42, 33.77, 42.86, 27.29, 31.99, 29.12, 29.48, 22.54, 31.57
#     ],
#     "Agro Specific": [
#         80.75, 79.58, 79.84, 72.12, 76.06, 75.03, 74.51, 52.02, 44.47, 42.00, 70.61
#     ],
#     "Remote Sensing Understanding": [
#         22.85, 16.93, 22.28, 17.07, 14.95, 10.72, 6.67, 12.56, 5.36, 8.33, 27.53
#     ]
# }

df = pd.DataFrame(data)
df['Average Score'] = df.iloc[:, 1:].mean(axis=1).round(2)
df = df[['Method', 'Average Score'] + [col for col in df.columns if col not in ['Method', 'Average Score']]]

def display_data():
    return df

with gr.Blocks() as demo:
    gr.Markdown("![camel icon](https://cdn-uploads.huggingface.co/production/uploads/656864e12d73834278a8dea7/n-XfVKd1xVywH_vgPyJyQ.png)", elem_id="camel-icon")  # Replace with actual camel icon URL
    gr.Markdown("# **CAMEL-Bench: Model Performance Across Vision Understanding Tasks**")
    gr.Markdown("""
    This table shows the performance of different models across various tasks including OCR, chart understanding, video, medical imaging, and more. 
    """)
    with gr.Tabs(elem_classes="tab-buttons") as tabs:
        with gr.TabItem("🏅 LLM Leaderboard", elem_id="llm-benchmark-tab-table", id=0):
            # with gr.Row():
                # with gr.Column():
            gr.Dataframe(value=df, label="CAMEL-Bench Model Performance", interactive=False)

        with gr.TabItem("📤 How to Submit", elem_id="submission-tab", id=1):
            gr.Markdown("""
            ## Submission Instructions

            To contribute your model's results to the CAMEL-Bench leaderboard:

            - **Via GitHub Pull Request**: 
              - Use [this evaluation script](https://github.com/mbzuai-oryx/Camel-Bench/blob/main/scripts/eval_qwen.py) to test your model and generate results.
              - Create a pull request in the CAMEL-Bench GitHub repository with your results.

            - **Via Email**:
              - Send your results to **[email protected]**, and we’ll add them to the leaderboard for you.

            **We look forward to seeing your contributions!**
            """)

demo.launch()