OpenDriveLab-org commited on
Commit
25a8d55
·
verified ·
1 Parent(s): 0425ee5

Release SparseVideoNav dataset

Browse files
.gitattributes CHANGED
@@ -58,3 +58,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
61
+ bvn/annotations.json filter=lfs diff=lfs merge=lfs -text
62
+ bvn/data.jsonl filter=lfs diff=lfs merge=lfs -text
63
+ ifn/annotations.json filter=lfs diff=lfs merge=lfs -text
64
+ ifn/data.jsonl filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-nc-sa-4.0
3
+ language:
4
+ - en
5
+ pretty_name: SparseVideoNav Datasets
6
+ task_categories:
7
+ - robotics
8
+ - visual-question-answering
9
+ tags:
10
+ - embodied-ai
11
+ - vision-language-navigation
12
+ - robot-navigation
13
+ - video
14
+ - trajectory
15
+ - sparsevideonav
16
+ - opendrivelab
17
+ configs:
18
+ - config_name: bvn
19
+ data_files:
20
+ - split: train
21
+ path: bvn/data.jsonl
22
+ - config_name: ifn
23
+ data_files:
24
+ - split: train
25
+ path: ifn/data.jsonl
26
+ ---
27
+
28
+ # SparseVideoNav Datasets
29
+
30
+ This repository contains the real-world navigation datasets released with [OpenDriveLab/SparseVideoNav](https://github.com/OpenDriveLab/SparseVideoNav):
31
+
32
+ - **BVN**: Beyond-the-View Navigation.
33
+ - **IFN**: Instruction-Following Navigation.
34
+
35
+ Project links:
36
+
37
+ - Project page: https://opendrivelab.com/SparseVideoNav
38
+ - GitHub: https://github.com/OpenDriveLab/SparseVideoNav
39
+ - Paper: https://arxiv.org/abs/2602.05827
40
+
41
+ ## Dataset Summary
42
+
43
+ SparseVideoNav studies real-world vision-language navigation with sparse future video generation. The datasets contain language instructions, RGB frame sequences, and low-level navigation actions. The number of actions matches the number of RGB frames for every released episode.
44
+
45
+ | Subset | Episodes | RGB frames | Task |
46
+ | --- | ---: | ---: | --- |
47
+ | `bvn` | 5,417 | 695,373 | Beyond-the-View Navigation |
48
+ | `ifn` | 5,625 | 712,290 | Instruction-Following Navigation |
49
+
50
+ ## Repository Structure
51
+
52
+ Images are stored in compressed tar shards to avoid hundreds of thousands of small files in the Hugging Face repository. Each shard preserves the original relative paths.
53
+
54
+ ```text
55
+ .
56
+ ├── README.md
57
+ ├── bvn/
58
+ │ ├── annotations.json
59
+ │ ├── data.jsonl
60
+ │ ├── merge_info.json
61
+ │ ├── shard_manifest.jsonl
62
+ │ └── shards/
63
+ │ ├── bvn-00000.tar.zst
64
+ │ └── ...
65
+ └── ifn/
66
+ ├── annotations.json
67
+ ├── data.jsonl
68
+ ├── merge_info.json
69
+ ├── shard_manifest.jsonl
70
+ └── shards/
71
+ ├── ifn-00000.tar.zst
72
+ └── ...
73
+ ```
74
+
75
+ Each line in `bvn/data.jsonl` or `ifn/data.jsonl` is an episode-level JSON object. Important fields:
76
+
77
+ - `episode_id`: episode identifier.
78
+ - `instruction` / `instructions`: natural-language navigation instruction.
79
+ - `actions`: per-frame low-level action sequence with `dx`, `dy`, and `dyaw`.
80
+ - `frame_files`: RGB frame paths inside the shard after extraction.
81
+ - `shard`: shard file containing the episode frames.
82
+
83
+ Example:
84
+
85
+ ```json
86
+ {
87
+ "dataset": "ifn",
88
+ "episode_id": "36e980e5a1ff4817",
89
+ "instruction": "please go along with the rail until you are near by a red cone.",
90
+ "num_frames": 177,
91
+ "num_actions": 177,
92
+ "rgb_dir": "images/0e0ec1bdcc61a4ef/rgb",
93
+ "frame_files": ["images/0e0ec1bdcc61a4ef/rgb/001.jpg"],
94
+ "shard": "ifn/shards/ifn-00000.tar.zst"
95
+ }
96
+ ```
97
+
98
+ ## Usage
99
+
100
+ Load episode metadata with Hugging Face Datasets:
101
+
102
+ ```python
103
+ from datasets import load_dataset
104
+
105
+ bvn = load_dataset("OpenDriveLab/SparseVideoNav", "bvn")
106
+ ifn = load_dataset("OpenDriveLab/SparseVideoNav", "ifn")
107
+ ```
108
+
109
+ Download and inspect shards:
110
+
111
+ ```bash
112
+ tar -I zstd -tf ifn/shards/ifn-00000.tar.zst | head
113
+ tar -I zstd -xf ifn/shards/ifn-00000.tar.zst
114
+ ```
115
+
116
+ After extraction, `frame_files` resolve to paths such as:
117
+
118
+ ```text
119
+ images/<episode_dir>/rgb/001.jpg
120
+ ```
121
+
122
+ ## License
123
+
124
+ The dataset is released under CC BY-NC-SA 4.0.
125
+
126
+ ## Citation
127
+
128
+ ```bibtex
129
+ @article{zhang2026sparse,
130
+ title={Sparse Video Generation Propels Real-World Beyond-the-View Vision-Language Navigation},
131
+ author={Zhang, Hai and Liang, Siqi and Chen, Li and Li, Yuxian and Xu, Yukuan and Zhong, Yichao and Zhang, Fu and Li, Hongyang},
132
+ journal={arXiv preprint arXiv:2602.05827},
133
+ year={2026}
134
+ }
135
+ ```
bvn/annotations.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92ba66d783f5e8f6d7657c106cb179f3658f54e41973963b74d00782ff3e3d9e
3
+ size 88454070
bvn/data.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4efb1dfe75a1acbf9c522441d151e84cce07c60fd7d3b3956d25570ce317a509
3
+ size 91532428
bvn/merge_info.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "category": "bvn",
3
+ "sources": [
4
+ "train_bvn",
5
+ "val_bvn"
6
+ ],
7
+ "source_annotation_counts": {
8
+ "train": 7475,
9
+ "val": 256
10
+ },
11
+ "deduplicated_by": "video",
12
+ "duplicates_removed": 71,
13
+ "merged_annotation_count": 7660,
14
+ "media_storage": "hard_links"
15
+ }
bvn/shard_manifest.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
bvn/shards/bvn-00000.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c6817badf1e4d897d7130b367332a6b582f58567e4109c5961970b1bfc75760
3
+ size 2089245353
bvn/shards/bvn-00001.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:522c18c16052416c9354429cd8856a4478e938b303a931260acd2eae75c8a6c2
3
+ size 2089144609
bvn/shards/bvn-00002.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2e7843be8f1b8b6f4f0662f089ef723b3f6ceeeed5846a462c9c1ca29657fe1
3
+ size 2090299990
bvn/shards/bvn-00003.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc820d1ee285e69435cffa036ba0b6029a73fc575fe147ced24b8072ff88444f
3
+ size 2089431677
bvn/shards/bvn-00004.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93a2e1648fa05b8f22c1b97d70a7de266c2df004a6a086cba1781c715951b6f6
3
+ size 2090463753
bvn/shards/bvn-00005.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c1f2c27c155d23122b200a52fa36ed832206a4cd84209c00c03e9371e5a2a04
3
+ size 2091368489
bvn/shards/bvn-00006.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b7a85c0755ff3dab4faee038b4cd6fa882fcb89682b179acfa8e1434d9de34d0
3
+ size 2093230366
bvn/shards/bvn-00007.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0175655d953b29d04a701ba538eeb98cf57fd612cd94923c4b648cdb75d94ba9
3
+ size 446456726
ifn/annotations.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d463d715bf25c59bae5da84606ce1610c50d561a1892a5a9231e39d4a5b45c1e
3
+ size 90633708
ifn/data.jsonl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d100c8794642645ea6a2dd70c822f96e65018634fc9e165a5dbff4dbffc88c73
3
+ size 93838245
ifn/merge_info.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "category": "ifn",
3
+ "sources": [
4
+ "train_ifn",
5
+ "val_ifn"
6
+ ],
7
+ "source_annotation_counts": {
8
+ "train": 5376,
9
+ "val": 263
10
+ },
11
+ "deduplicated_by": "video",
12
+ "duplicates_removed": 13,
13
+ "merged_annotation_count": 5626,
14
+ "media_storage": "hard_links"
15
+ }
ifn/shard_manifest.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
ifn/shards/ifn-00000.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eba64794e3b76bcdab6815c5fd939c9dd392a005d629c29dca060f070eb8adfd
3
+ size 2092533721
ifn/shards/ifn-00001.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f91a0fc7f1be970bb4c2e72f16435df07f05feab653b9e37464b57b372ccb311
3
+ size 2089041896
ifn/shards/ifn-00002.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:65064016a782a11b0a62735df2825bb2d18370a9419a7a2fd1d17704bb9cc86e
3
+ size 2088625389
ifn/shards/ifn-00003.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7617efb133b877cfbf49e8edc2277805e3194f0c88b473bcb342c56355c2a247
3
+ size 2092873186
ifn/shards/ifn-00004.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:814dbaaec64ae90bac8dc5bb0c5f1c9e55e218109d0367974fbd0e57f2fc5315
3
+ size 2089270756
ifn/shards/ifn-00005.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fba9371ae1267583d84f5596569def7704973536e5bb7900df5ab67311915078
3
+ size 2089015728
ifn/shards/ifn-00006.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:42025cc433fe237c3ce069022ee658a719e19041392a99cfc2449e99b70c55c1
3
+ size 2093474869
ifn/shards/ifn-00007.tar.zst ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4888445c64c4e601d64f2ddc237a20ed17f0f3a8f9e8b35d7a56acb1bc7a5dc
3
+ size 1031294189