Update README.md
Browse filesinit spatialeval description
README.md
CHANGED
@@ -73,3 +73,60 @@ configs:
|
|
73 |
path: vtqa/test-*
|
74 |
---
|
75 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
73 |
path: vtqa/test-*
|
74 |
---
|
75 |
|
76 |
+
## 🤔 About SpatialEval
|
77 |
+
SpatialEval is a comprehensive benchmark for evaluating spatial intelligence in LLMs and VLMs across four key dimensions:
|
78 |
+
- Spatial relationships
|
79 |
+
- Positional understanding
|
80 |
+
- Object counting
|
81 |
+
- Navigation
|
82 |
+
|
83 |
+
### Benchmark Tasks
|
84 |
+
1. **Spatial-Map**: Understanding spatial relationships between objects in map-based scenarios
|
85 |
+
2. **Maze-Nav**: Testing navigation through complex environments
|
86 |
+
3. **Spatial-Grid**: Evaluating spatial reasoning within structured environments
|
87 |
+
4. **Spatial-Real**: Assessing real-world spatial understanding
|
88 |
+
|
89 |
+
Each task supports three input modalities:
|
90 |
+
- Text-only (TQA)
|
91 |
+
- Vision-only (VQA)
|
92 |
+
- Vision-Text (VTQA)
|
93 |
+
|
94 |
+
![spatialeval_task.png](https://cdn-uploads.huggingface.co/production/uploads/651651f5d93a51ceda3021c3/kpjld6-HCg5LXhO9Ju6-Q.png)
|
95 |
+
|
96 |
+
## 📌 Quick Links
|
97 |
+
Project Page: https://spatialeval.github.io/
|
98 |
+
Paper: https://arxiv.org/pdf/2406.14852
|
99 |
+
Code: https://github.com/jiayuww/SpatialEval
|
100 |
+
Talk: https://neurips.cc/virtual/2024/poster/94371
|
101 |
+
|
102 |
+
## 🚀 Quick Start
|
103 |
+
|
104 |
+
### 📍 Load Dataset
|
105 |
+
|
106 |
+
SpatialEval provides three input modalities—TQA (Text-only), VQA (Vision-only), and VTQA (Vision-text)—across four tasks: Spatial-Map, Maze-Nav, Spatial-Grid, and Spatial-Real. Each modality and task is easily accessible via Hugging Face. Ensure you have installed the [packages](https://huggingface.co/docs/datasets/en/quickstart):
|
107 |
+
|
108 |
+
```python
|
109 |
+
from datasets import load_dataset
|
110 |
+
|
111 |
+
tqa = load_dataset("MilaWang/SpatialEval", "tqa", split="test")
|
112 |
+
vqa = load_dataset("MilaWang/SpatialEval", "vqa", split="test")
|
113 |
+
vtqa = load_dataset("MilaWang/SpatialEval", "vtqa", split="test")
|
114 |
+
```
|
115 |
+
|
116 |
+
## ⭐ Citation
|
117 |
+
|
118 |
+
If you find our work helpful, please consider citing our paper 😊
|
119 |
+
|
120 |
+
```
|
121 |
+
@inproceedings{wang2024spatial,
|
122 |
+
title={Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models},
|
123 |
+
author={Wang, Jiayu and Ming, Yifei and Shi, Zhenmei and Vineet, Vibhav and Wang, Xin and Li, Yixuan and Joshi, Neel},
|
124 |
+
booktitle={The Thirty-Eighth Annual Conference on Neural Information Processing Systems},
|
125 |
+
year={2024}
|
126 |
+
}
|
127 |
+
```
|
128 |
+
|
129 |
+
## 💬 Questions
|
130 |
+
Have questions? We're here to help!
|
131 |
+
- Open an issue in the github repository
|
132 |
+
- Contact us through the channels listed on our project page
|