Update README.md

README.md

@@ -54,7 +54,7 @@ dataset_info:
     num_bytes: 17876685412.11577
     num_examples: 502
   - name: test
-    num_bytes: 17565562113.0
+    num_bytes: 17565562113
     num_examples: 500
   download_size: 369722031
   dataset_size: 35442247525.11577
@@ -65,4 +65,111 @@ configs:
     path: data/validation-*
   - split: test
     path: data/test-*
+tags:
+- image
+- text
+- navigation
+pretty_name: NaviTrace
+license: cc-by-4.0
+task_categories:
+- visual-question-answering
+- robotics
+language:
+- en
+size_categories:
+- 1K<n<10K
 ---
+
+<style>
+    /* Headings */
+    h1 {
+        text-align: center;
+        font-size: 4.5rem !important;
+        font-weight: 500;
+        margin-top: 1rem;
+        margin-bottom: 2rem;
+    }
+    
+    /* Links container */
+    .links-container {
+        text-align: center;
+        margin-bottom: 3rem;
+        font-size: 1.1rem;
+    }
+    
+    .links-container a {
+        margin: 0 1rem;
+        text-decoration: none;
+        color: #3b82f6;
+        font-weight: 600;
+        transition: color 0.3s;
+    }
+    
+    .links-container a:hover {
+        color: #1e3a8a;
+    }
+</style>
+
+<div align="center">
+
+<h1><b>NaviTrace</b></h1>
+
+<div class="links-container">
+    <a href="https://leggedrobotics.github.io/navitrace_webpage/" target="_blank">
+        🏠 Project
+    </a>
+    <a href="#" target="_blank">
+        📄 Paper
+    </a>
+    <a href="#" target="_blank">
+        💻 Code
+    </a>
+    <a href="https://huggingface.co/spaces/leggedrobotics/navitrace_leaderboard" target="_blank">
+        🏆 Leaderboard
+    </a>
+</div>
+
+<img src="https://leggedrobotics.github.io/navitrace_webpage/static/images/Figure_1.png" alt="NaviTrace Overview" width="800px">
+
+***NaviTrace** is a novel VQA benchmark for VLMs that evaluates models on their embodiment-specific understanding of navigation across challenging real-world scenarios.*
+
+</div>
+
+## Key Features
+
+- ✏️ **Core Task:** Given a real-world image in first-person perspective, a language instruction, and an embodiment type, models should predict a 2D navigation path in image space that solves the instruction.
+- 🤖 **Embodiments:** Four embodiment types capturing distinct physical and spatial constraints (human, legged robot, wheeled robot, or bicycle).
+- 📏 **Scale:** 1,000 diverse real-world scenarios and over 3,000 expert-annotated traces.
+- ⚖️ **Splits:**
+  - Validation split (50%) for model fine-tuning.
+  - Test split (50%) with hidden ground-truths for public leaderboard evaluation.
+- 🔎 **Annotation Quality:** All images and traces manually collected and labeled by human experts.
+- 🏅 **Evaluation Metric:** Semantic-aware Trace Score, combining Dynamic Time Warping distance, Goal endpoint error, and Embodiment-conditioned semantic penalties.
+
+## Uses
+
+### Run Benchmark
+
+TODO See notbook
+
+### Model Training
+
+TODO
+
+## Structure
+
+only suitable embodiments
+
+## Citation
+
+If you find this dataset helpful for your work, please cite us with:
+
+**BibTeX:**
+
+```
+TODO
+```
+
+## Contact
+
+TODO