satishvutukuru commited on
Commit
d70b0e9
·
verified ·
1 Parent(s): 5c1b8ef

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +84 -24
README.md CHANGED
@@ -1,28 +1,88 @@
1
  ---
2
- license: mit
3
- configs:
4
- - config_name: default
5
- data_files:
6
- - split: train
7
- path: data/train-*
 
 
 
 
 
 
 
8
  dataset_info:
9
  features:
10
- - name: id
11
- dtype: string
12
- - name: image
13
- dtype: image
14
- - name: category
15
- dtype: string
16
- - name: question
17
- dtype: string
18
- - name: choices
19
- list: string
20
- - name: answer
21
- dtype: string
22
- splits:
23
- - name: train
24
- num_bytes: 21337420
25
- num_examples: 50
26
- download_size: 21285474
27
- dataset_size: 21337420
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
28
  ---
 
1
  ---
2
+ license: cc-by-4.0
3
+ task_categories:
4
+ - multiple-choice
5
+ language:
6
+ - en
7
+ tags:
8
+ - autonomous-driving
9
+ - spatial-reasoning
10
+ - 3d-reasoning
11
+ - vqa
12
+ pretty_name: Driving 3D Spatial Reasoning
13
+ size_categories:
14
+ - n<1K
15
  dataset_info:
16
  features:
17
+ - name: id
18
+ dtype: string
19
+ - name: image
20
+ dtype: image
21
+ - name: category
22
+ dtype: string
23
+ - name: question
24
+ dtype: string
25
+ - name: choices
26
+ sequence: string
27
+ - name: answer
28
+ dtype: string
29
+ ---
30
+
31
+ # Open Spatial Reasoning
32
+
33
+ A multiple-choice dataset of **spatial reasoning questions and answers** for evaluating **3D spatial reasoning** from **single driving images**. Each image contains numbered bounding boxes referencing objects in the scene, and each question probes a model's ability to reconstruct the real 3D scene rather than rely on flat-image shortcuts (e.g. "lower in the frame = closer", "bigger box = nearer").
34
+
35
+ ## Dataset Description
36
+
37
+ Frontier vision-language models often answer these questions correctly by luck while reasoning incorrectly, leaning on pixel-layout heuristics that break down on elevated roads, slopes, curves, and intersections. This dataset is designed to surface that failure mode by requiring metric 3D reasoning about distance, lateral position, ordering, and heading.
38
+
39
+ Each sample pairs a driving-scene image with a question, four answer choices, and the correct answer letter.
40
+
41
+ The images were collected by autonomous vehicles operated by **Plus AI**.
42
+
43
+ ### Data Fields
44
+
45
+ | Field | Type | Description |
46
+ |---|---|---|
47
+ | `id` | `string` | Unique question identifier (e.g. `q0002`) |
48
+ | `image` | `image` | The driving image with numbered bounding boxes |
49
+ | `category` | `string` | The reasoning task type (see categories below) |
50
+ | `question` | `string` | The full question, including the reasoning protocol |
51
+ | `choices` | `list[string]` | Four answer options, prefixed `A.`–`D.` |
52
+ | `answer` | `string` | The correct answer letter (`A`, `B`, `C`, or `D`) |
53
+
54
+ ### Question Categories
55
+
56
+ The dataset spans several spatial-reasoning task types, including:
57
+
58
+ | Category | What it tests |
59
+ |---|---|
60
+ | `identify_distance_long` | Estimate the absolute distance to an object (binned 0–20m / 20–50m / 50–80m / 80m+) |
61
+ | `relative_distance_long` | Estimate the 3D separation between two objects |
62
+ | `pick_closer` | Decide which of two objects is closer to the ego vehicle |
63
+ | `identify_rightmost` | Identify the object furthest to the right in true 3D space |
64
+ | `order_leftmost` | Order several objects left-to-right in 3D space |
65
+ | `identify_position` | Classify an object's position relative to ego (e.g. ahead-left, behind-right) |
66
+ | `identify_heading` | Determine an object's heading using clock directions (12 = forward, 3 = right) |
67
+
68
+ ## Authors
69
+
70
+ Anurag Ganguli, [Anshuman Lall](https://huggingface.co/anshuman-reasoncore), Abhishek Bhatia, Xiangyu Gao, Joe Yuan, [Satish Vutukuru](https://huggingface.co/satishvutukuru), [Geoff Wolfe](https://huggingface.co/geoffwolfeRC)
71
+
72
+ ## Citation
73
+
74
+ If you use this dataset, please cite it:
75
+
76
+ ```bibtex
77
+ @misc{driving_3d_spatial_reasoning,
78
+ title = {Open Spatial Reasoning},
79
+ author = {Anurag Ganguli, Anshuman Lall, Abhishek Bhatia, Xiangyu Gao, Joe Yuan, Satish Vutukuru, Geoff Wolfe},
80
+ year = {2026},
81
+ howpublished = {\url{https://huggingface.co/datasets/your-username/your-dataset-name}}
82
+ }
83
+ ```
84
+
85
+ ## License
86
+
87
+ Released under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). Images were collected by autonomous vehicles operated by Plus AI.
88
  ---