Initial Commit

README.md

@@ -0,0 +1,61 @@
+---
+title: Mars Vision Leaderboard
+emoji: 🚀
+colorFrom: red
+colorTo: orange
+sdk: gradio
+sdk_version: 4.0.0
+app_file: app.py
+pinned: false
+---
+
+# Mars Vision Leaderboard 🚀
+
+A modern leaderboard for comparing state-of-the-art computer vision models across different tasks: Classification, Segmentation, and Object Detection.
+
+## Features
+
+- **Three Task Categories**: Classification, Segmentation, and Object Detection
+- **Interactive Filters**:
+  - Dataset selection with checkboxes
+  - Metrics selection with checkboxes
+  - Model type/organization filtering
+- **Real-time Updates**: Table updates dynamically based on selected filters
+- **Clean UI**: Modern design inspired by leading HuggingFace leaderboards
+
+## How to Use
+
+1. Select a tab for the task you're interested in (Classification, Segmentation, or Object Detection)
+2. Use the filter boxes to select:
+   - Which datasets to display
+   - Which metrics to show
+   - Which model types/organizations to include
+3. The leaderboard table updates automatically based on your selections
+
+## Local Development
+
+```bash
+# Install dependencies
+pip install -r requirements.txt
+
+# Run the app
+python app.py
+```
+
+## Customization
+
+To add your own data, modify the data dictionaries in `app.py`:
+- `CLASSIFICATION_DATA`
+- `SEGMENTATION_DATA`
+- `DETECTION_DATA`
+
+Each entry should have:
+- `model`: Model name
+- `organization`: Organization/creator
+- `dataset`: Dataset name
+- `metric`: Metric name
+- `value`: Numeric value
+
+## License
+
+MIT

app.py

@@ -0,0 +1,221 @@
+import gradio as gr
+import pandas as pd
+
+CLASSIFICATION_DATA = {
+    "Model": [
+        "ResNet-50", "ViT-Base", "Swin-T", "InceptionV3", "SqueezeNet",
+        "ResNet-50", "ViT-Base", "Swin-T", "InceptionV3", "SqueezeNet",
+        "ResNet-50", "ViT-Base", "Swin-T", "InceptionV3", "SqueezeNet",
+    ],
+    "Dataset": [
+        "DoMars16", "DoMars16", "DoMars16", "DoMars16", "DoMars16",
+        "Atmospheric Dust", "Atmospheric Dust", "Atmospheric Dust", "Atmospheric Dust", "Atmospheric Dust",
+        "Martian Frost", "Martian Frost", "Martian Frost", "Martian Frost", "Martian Frost",
+    ],
+    "Accuracy": [
+        92.5, 94.2, 95.8, 93.1, 89.7,
+        88.3, 90.1, 91.5, 89.8, 87.2,
+        85.6, 87.9, 88.4, 86.7, 84.3,
+    ],
+    "F1-Score": [
+        91.8, 93.5, 94.9, 92.4, 88.6,
+        87.5, 89.2, 90.7, 88.9, 86.3,
+        84.8, 86.9, 87.5, 85.8, 83.4,
+    ],
+}
+
+DETECTION_DATA = {
+    "Model": [
+        "Faster R-CNN", "YOLOv5", "DETR", "RetinaNet", "SSD",
+        "Faster R-CNN", "YOLOv5", "DETR", "RetinaNet", "SSD",
+        "Faster R-CNN", "YOLOv5", "DETR", "RetinaNet", "SSD",
+    ],
+    "Dataset": [
+        "Mars Crater", "Mars Crater", "Mars Crater", "Mars Crater", "Mars Crater",
+        "Rover Component", "Rover Component", "Rover Component", "Rover Component", "Rover Component",
+        "Geological Feature", "Geological Feature", "Geological Feature", "Geological Feature", "Geological Feature",
+    ],
+    "mAP": [
+        78.5, 80.2, 82.1, 79.3, 77.8,
+        75.6, 77.3, 78.9, 76.7, 75.1,
+        73.4, 75.1, 76.7, 74.5, 73.0,
+    ],
+    "IoU": [
+        0.72, 0.74, 0.76, 0.73, 0.71,
+        0.69, 0.71, 0.73, 0.70, 0.68,
+        0.67, 0.69, 0.71, 0.68, 0.67,
+    ],
+}
+
+SEGMENTATION_DATA = {
+    "Model": [
+        "U-Net", "DeepLabV3+", "Mask R-CNN", "SegFormer", "HRNet",
+        "U-Net", "DeepLabV3+", "Mask R-CNN", "SegFormer", "HRNet",
+        "U-Net", "DeepLabV3+", "Mask R-CNN", "SegFormer", "HRNet",
+    ],
+    "Dataset": [
+        "Mars Terrain", "Mars Terrain", "Mars Terrain", "Mars Terrain", "Mars Terrain",
+        "Dust Storm", "Dust Storm", "Dust Storm", "Dust Storm", "Dust Storm",
+        "Geological Feature", "Geological Feature", "Geological Feature", "Geological Feature", "Geological Feature",
+    ],
+    "Dice Score": [
+        0.85, 0.87, 0.88, 0.86, 0.84,
+        0.82, 0.84, 0.85, 0.83, 0.82,
+        0.81, 0.83, 0.84, 0.82, 0.81,
+    ],
+    "IoU": [
+        0.74, 0.76, 0.78, 0.75, 0.73,
+        0.70, 0.72, 0.74, 0.71, 0.70,
+        0.68, 0.70, 0.72, 0.69, 0.68,
+    ],
+}
+
+
+def filter_and_search(df, search, datasets, models, columns):
+    filtered = df.copy()
+
+    if search:
+        mask = filtered.apply(lambda row: row.astype(str).str.contains(search, case=False).any(), axis=1)
+        filtered = filtered[mask]
+
+    if datasets:
+        filtered = filtered[filtered["Dataset"].isin(datasets)]
+    if models:
+        filtered = filtered[filtered["Model"].isin(models)]
+
+    if columns:
+        display_cols = [col for col in columns if col in filtered.columns]
+        filtered = filtered[display_cols]
+
+    return filtered
+
+
+def build_tab(data, name):
+    df = pd.DataFrame(data)
+    datasets = sorted(df["Dataset"].unique().tolist())
+    models = sorted(df["Model"].unique().tolist())
+    metric_cols = [col for col in df.columns if col not in ["Model", "Dataset"]]
+    all_cols = ["Model", "Dataset"] + metric_cols
+
+    with gr.TabItem(name, elem_id="llm-benchmark-tab-table"):
+        with gr.Row():
+            with gr.Column(scale=4):
+                search_bar = gr.Textbox(
+                    label="Search",
+                    placeholder="Separate multiple queries with ';'",
+                    elem_id="search-bar"
+                )
+
+                col_selector = gr.CheckboxGroup(
+                    choices=all_cols,
+                    value=all_cols,
+                    label="Select Columns to Display:",
+                    elem_classes="column-select"
+                )
+
+                table = gr.Dataframe(
+                    value=df,
+                    wrap=True,
+                    interactive=False,
+                    elem_id="leaderboard-table"
+                )
+
+            with gr.Column(scale=1):
+                gr.Markdown("**Model types**")
+                model_filter = gr.CheckboxGroup(
+                    choices=models,
+                    value=models,
+                    label="",
+                    elem_classes="filter-group"
+                )
+
+                gr.Markdown("**Datasets**")
+                dataset_filter = gr.CheckboxGroup(
+                    choices=datasets,
+                    value=datasets,
+                    label="",
+                    elem_classes="filter-group"
+                )
+
+        def update(search, ds, md, cols):
+            return filter_and_search(df, search, ds, md, cols)
+
+        search_bar.change(update, [search_bar, dataset_filter, model_filter, col_selector], table)
+        dataset_filter.change(update, [search_bar, dataset_filter, model_filter, col_selector], table)
+        model_filter.change(update, [search_bar, dataset_filter, model_filter, col_selector], table)
+        col_selector.change(update, [search_bar, dataset_filter, model_filter, col_selector], table)
+
+
+custom_css = """
+.markdown-text {
+    font-size: 16px !important;
+}
+
+#leaderboard-table {
+    margin-top: 15px;
+}
+
+#search-bar-table-box > div:first-child {
+    background: none;
+    border: none;
+}
+
+#search-bar {
+    padding: 0px;
+}
+
+table td:first-child,
+table th:first-child {
+    max-width: 400px;
+    overflow: auto;
+    white-space: nowrap;
+}
+
+.tab-buttons button {
+    font-size: 20px;
+}
+
+.filter-group {
+    margin-bottom: 1em;
+}
+
+.filter-group label {
+    font-size: 14px;
+}
+
+.column-select {
+    margin-bottom: 1.5em;
+}
+
+.column-select label {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 0.5em;
+}
+
+/* Styling for the column select checkboxes to display in rows */
+.column-select label > span {
+    display: inline-flex;
+    align-items: center;
+}
+"""
+
+TITLE = """<h1 align="center" id="space-title">Mars Vision Leaderboard</h1>"""
+
+INTRO = """
+A comprehensive benchmark for evaluating computer vision models on Mars-specific datasets.
+This leaderboard tracks model performance across three key tasks: classification, segmentation, and object detection.
+"""
+
+demo = gr.Blocks(css=custom_css, title="Mars Vision Leaderboard")
+with demo:
+    gr.HTML(TITLE)
+    gr.Markdown(INTRO, elem_classes="markdown-text")
+
+    with gr.Tabs(elem_classes="tab-buttons"):
+        build_tab(CLASSIFICATION_DATA, "🏅 Classification")
+        build_tab(SEGMENTATION_DATA, "🏅 Segmentation")
+        build_tab(DETECTION_DATA, "🏅 Object Detection")
+
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -0,0 +1,2 @@
+gradio>=4.0.0
+pandas>=2.0.0