Hugging Face's logo

Datasets:
Claimentine
/
CoVT_scaleup

Tasks:
Visual Question Answering
Image-to-Text
Modalities:
Tabular
Size:
< 1K
Tags:
chain-of-thought
depth-estimation
edge-detection
segmentation
webdataset
Libraries:
WebDataset
Dataset card Data Studio Files
xet
 Community
1
Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
shard_idx
int64
0
296
n_records
int64
1.79k
2k
n_images
int64
5.37k
7.83k
n_missing
int64
0
0
done
bool
1 class
0
2,000
7,474
0
true
1
2,000
7,451
0
true
2
2,000
7,479
0
true
3
2,000
7,478
0
true
4
2,000
7,483
0
true
5
2,000
7,471
0
true
6
2,000
7,472
0
true
7
2,000
7,500
0
true
8
2,000
7,460
0
true
9
2,000
7,485
0
true
10
2,000
7,501
0
true
11
2,000
7,437
0
true
12
2,000
7,454
0
true
13
2,000
7,541
0
true
14
2,000
7,491
0
true
15
2,000
6,744
0
true
16
2,000
6,722
0
true
17
2,000
6,720
0
true
18
2,000
6,700
0
true
19
2,000
6,864
0
true
20
2,000
6,946
0
true
21
2,000
6,933
0
true
22
2,000
6,222
0
true
23
2,000
6,000
0
true
24
2,000
6,000
0
true
25
2,000
6,000
0
true
26
2,000
6,000
0
true
27
2,000
6,000
0
true
28
2,000
6,000
0
true
29
2,000
6,000
0
true
30
2,000
6,000
0
true
31
2,000
6,000
0
true
32
2,000
6,000
0
true
33
2,000
6,000
0
true
34
2,000
6,000
0
true
35
2,000
6,000
0
true
36
2,000
6,000
0
true
37
2,000
6,000
0
true
38
2,000
6,000
0
true
39
2,000
6,000
0
true
40
2,000
6,000
0
true
41
2,000
6,000
0
true
42
2,000
6,000
0
true
43
2,000
6,000
0
true
44
2,000
6,000
0
true
45
2,000
6,000
0
true
46
2,000
6,000
0
true
47
2,000
6,000
0
true
48
2,000
6,000
0
true
49
2,000
6,000
0
true
50
2,000
6,000
0
true
51
2,000
6,000
0
true
52
2,000
6,000
0
true
53
2,000
6,000
0
true
54
2,000
6,000
0
true
55
2,000
6,000
0
true
56
2,000
6,000
0
true
57
2,000
7,040
0
true
58
2,000
7,205
0
true
59
2,000
7,279
0
true
60
2,000
7,471
0
true
61
2,000
7,574
0
true
62
2,000
7,582
0
true
63
2,000
7,598
0
true
64
2,000
7,579
0
true
65
2,000
7,598
0
true
66
2,000
7,618
0
true
67
2,000
7,583
0
true
68
2,000
7,611
0
true
69
2,000
7,592
0
true
70
2,000
7,573
0
true
71
2,000
7,589
0
true
72
2,000
7,564
0
true
73
2,000
7,587
0
true
74
2,000
7,585
0
true
75
2,000
7,335
0
true
76
2,000
7,542
0
true
77
2,000
7,553
0
true
78
2,000
7,567
0
true
79
2,000
7,559
0
true
80
2,000
7,543
0
true
81
2,000
7,544
0
true
82
2,000
7,576
0
true
83
2,000
7,561
0
true
84
2,000
7,591
0
true
85
2,000
7,565
0
true
86
2,000
7,572
0
true
87
2,000
7,579
0
true
88
2,000
7,494
0
true
89
2,000
6,961
0
true
90
2,000
6,899
0
true
91
2,000
6,864
0
true
92
2,000
6,846
0
true
93
2,000
6,707
0
true
94
2,000
6,690
0
true
95
2,000
6,750
0
true
96
2,000
6,891
0
true
97
2,000
6,897
0
true
98
2,000
7,025
0
true
99
2,000
6,972
0
true
End of preview. Expand in Data Studio

CoVT_scaleup

Chain-of-thought VQA data for training a model to generate auxiliary depth / edge / segmentation maps as intermediate reasoning steps inside a <think> block, e.g.:

<think> The depth map of the image is [depth map], The edge map of the image
is [edge map], ... The segmentation of the image is [seg map]. </think>
The answer is X

Companion code: Claimentine/CoVT_scaleup on GitHub (BAGEL fine-tuning pipeline, ThinkMorph-style boundary signaling — see that repo's README for the modeling side).

Stats

  • 593,790 records
  • 297 WebDataset shards (shards/shard-00000.tar … shards/shard-00296.tar)
  • 1,942,181 images total (original RGB photo + generated depth map + PiDiNet edge map +, for a subset of records, a SAM3 segmentation map)
  • ~332 GB total

Format

Each shard is a plain tar archive following the WebDataset convention: files sharing the same zero-padded 9-digit key belong to one sample.

000000000.json          # metadata + conversation for this sample
000000000.img0.png      # original RGB photo
000000000.img1.png      # depth map
000000000.img2.png      # PiDiNet edge map
000000000.img3.png      # SAM3 segmentation map (present for most, not all, records)
000000001.json
000000001.img0.png
...

{key}.json is the original CoT record with image rewritten to point at the in-tar filenames instead of absolute cluster paths:

{
  "id": "identity_177477",
  "image": ["000000000.img0.png", "000000000.img1.png", "000000000.img2.png", "000000000.img3.png"],
  "conversations": [
    {"from": "human", "value": "<image>\nQuestion: ..."},
    {"from": "gpt", "value": "<think> The depth map of the image is <image>, The edge map of the image is <image>, ... </think> The answer is ..."}
  ],
  "seg_coverage": 0.3558
}

image[0] is always the original RGB photo; image[1:] are the auxiliary maps in the same order they're referenced by the <image> tags inside the <think> block of the gpt turn. Some records only have 3 images (no segmentation map) — the number of <image> tags in the gpt turn always matches len(image) - 1.

Each shard also has a shard-XXXXX.manifest.json (record count / image count for that shard — packing-time bookkeeping, not needed to load the data).

Loading 

import webdataset as wds

url = "https://huggingface.co/datasets/claimentine/CoVT_scaleup/resolve/main/shards/shard-{00000..00296}.tar"
ds = wds.WebDataset(url).decode("pil")

for sample in ds:
    record = sample["json"]
    rgb = sample["img0.png"]
    depth = sample["img1.png"]
    ...

Or download shards directly and use the loader in the companion code repo (data/cot_dataset.py:CoTJSONLIterableDataset expects a plain jsonl + loose image files on disk — extract shards locally first if you want to train with that exact loader unmodified).

Provenance / license note

Source questions/images are drawn from a mixture of existing VQA datasets (TallyQA, CLEVR, A-OKVQA, ST-VQA, and others reachable via the record id prefix, e.g. cauldron/tallyqa/...), each carrying their own upstream license — this repo does not re-license them. Depth maps, PiDiNet edge maps, and SAM3 segmentation maps are generated/derived by this project. No explicit license is declared at the dataset-repo level; treat as research-use data and check the upstream source dataset's license before any other use.
Downloads last month
57
Number of rows:
297
Total file size:
356 GB