Datasets:
Add files using upload-large-folder tool
Browse files- README.md +33 -43
- data/benchmark-00000-of-00001.parquet +3 -0
- data/train-00000-of-00009.parquet +3 -0
- data/train-00001-of-00009.parquet +3 -0
- data/train-00002-of-00009.parquet +3 -0
- data/train-00003-of-00009.parquet +3 -0
- data/train-00004-of-00009.parquet +3 -0
- data/train-00005-of-00009.parquet +3 -0
- data/train-00006-of-00009.parquet +3 -0
- data/train-00007-of-00009.parquet +3 -0
- data/train-00008-of-00009.parquet +3 -0
README.md
CHANGED
|
@@ -18,65 +18,56 @@ tags:
|
|
| 18 |
- multimodal
|
| 19 |
size_categories:
|
| 20 |
- 10K<n<100K
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
---
|
| 22 |
|
| 23 |
# StructHub
|
| 24 |
|
| 25 |
**StructHub** is a large-scale dataset of **structured images paired with their editable
|
| 26 |
-
SVG / XML source code**,
|
| 27 |
-
|
| 28 |
chart, flowchart, architecture figure, UML diagram, schema, etc.) is paired with
|
| 29 |
-
**object-level** vector code that explicitly encodes text, shapes, topology, and layout
|
| 30 |
-
supporting supervised and reward-based training for models that "make structured images
|
| 31 |
-
editable again."
|
| 32 |
|
| 33 |
- 📄 Paper: *Back2Struct: Making Structured Images Editable Again*
|
| 34 |
- 🔗 Project page: https://pengyu965.github.io/Back2Struct.github.io/
|
| 35 |
- 🤖 Model: [`Pengyu965/Back2Struct-Image2SVG-7B`](https://huggingface.co/Pengyu965/Back2Struct-Image2SVG-7B)
|
| 36 |
|
| 37 |
-
##
|
| 38 |
|
| 39 |
-
|
|
| 40 |
-
|-------
|
| 41 |
-
|
|
| 42 |
-
|
|
| 43 |
-
| **NNArch** | 3,293 | 1,453 | 46.6% | 97 |
|
| 44 |
-
| **Total** | **84,121** | **73,128** | **86.9%** | **1,000** |
|
| 45 |
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
- **NNArch** — neural-network / system architecture diagrams (long, complex SVGs).
|
| 49 |
|
| 50 |
-
|
| 51 |
-
stable long-context generation.
|
| 52 |
-
|
| 53 |
-
## StructHub Benchmark (1,000 samples)
|
| 54 |
-
|
| 55 |
-
A held-out **1,000-sample** evaluation benchmark (569 StarDiag-HQ + 334 OpenDiag +
|
| 56 |
-
97 NNArch), split by ground-truth SVG length into three difficulty tiers:
|
| 57 |
-
|
| 58 |
-
| Tier | GT SVG length | # samples |
|
| 59 |
-
|------|---------------|----------:|
|
| 60 |
-
| Easy | ≤ 2,048 tokens | 333 |
|
| 61 |
-
| Medium | 2,048 – 4,096 tokens | 333 |
|
| 62 |
-
| Hard | > 4,096 tokens | 334 |
|
| 63 |
-
|
| 64 |
-
This enables evaluation across levels of structural complexity.
|
| 65 |
-
|
| 66 |
-
## Data fields
|
| 67 |
-
|
| 68 |
-
Each example contains:
|
| 69 |
|
| 70 |
| Field | Type | Description |
|
| 71 |
|-------|------|-------------|
|
| 72 |
-
| `id` | string | Unique example id (source-prefixed, e.g. `starvector_…`, `crawled_…`) |
|
| 73 |
-
| `
|
|
|
|
|
|
|
| 74 |
| `svg` | string | Ground-truth object-level SVG / XML code (target) |
|
| 75 |
-
| `source` | string | `StarDiag-HQ` \| `OpenDiag` \| `NNArch` |
|
| 76 |
-
| `num_tokens` | int | Token length of the ground-truth SVG |
|
| 77 |
-
| `difficulty` | string | `easy` \| `medium` \| `hard` (benchmark split only) |
|
| 78 |
|
| 79 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
|
| 81 |
## Usage
|
| 82 |
|
|
@@ -96,14 +87,13 @@ print(ex["svg"][:500]) # ground-truth SVG/XML
|
|
| 96 |
|
| 97 |
Predictions are scored with **image-level** metrics (render success rate, DINOv2 similarity,
|
| 98 |
LPIPS, SSIM between the rendered prediction and the input image) and a **text-level**
|
| 99 |
-
GPTScore over the SVG code
|
| 100 |
-
paper for the render-gated evaluation protocol.
|
| 101 |
|
| 102 |
## Licensing
|
| 103 |
|
| 104 |
StructHub aggregates multiple sources, including crawled open-domain graphics. Per-source
|
| 105 |
terms may differ; the dataset is released for **research use**. Please review and respect the
|
| 106 |
-
licenses of the original content. *(Confirm and update this field before
|
| 107 |
|
| 108 |
## Citation
|
| 109 |
|
|
|
|
| 18 |
- multimodal
|
| 19 |
size_categories:
|
| 20 |
- 10K<n<100K
|
| 21 |
+
configs:
|
| 22 |
+
- config_name: default
|
| 23 |
+
data_files:
|
| 24 |
+
- split: train
|
| 25 |
+
path: data/train-*.parquet
|
| 26 |
+
- split: benchmark
|
| 27 |
+
path: data/benchmark-*.parquet
|
| 28 |
---
|
| 29 |
|
| 30 |
# StructHub
|
| 31 |
|
| 32 |
**StructHub** is a large-scale dataset of **structured images paired with their editable
|
| 33 |
+
SVG / XML source code**, for the *image-to-SVG* task introduced in
|
| 34 |
+
**Back2Struct: Making Structured Images Editable Again**. Each raster image (a diagram,
|
| 35 |
chart, flowchart, architecture figure, UML diagram, schema, etc.) is paired with
|
| 36 |
+
**object-level** vector code that explicitly encodes text, shapes, topology, and layout.
|
|
|
|
|
|
|
| 37 |
|
| 38 |
- 📄 Paper: *Back2Struct: Making Structured Images Editable Again*
|
| 39 |
- 🔗 Project page: https://pengyu965.github.io/Back2Struct.github.io/
|
| 40 |
- 🤖 Model: [`Pengyu965/Back2Struct-Image2SVG-7B`](https://huggingface.co/Pengyu965/Back2Struct-Image2SVG-7B)
|
| 41 |
|
| 42 |
+
## Splits
|
| 43 |
|
| 44 |
+
| Split | Description | # examples |
|
| 45 |
+
|-------|-------------|-----------:|
|
| 46 |
+
| `train` | Full StructHub training set (all sources, all token lengths), with the benchmark held out | 86,102 |
|
| 47 |
+
| `benchmark` | Held-out StructHub evaluation benchmark | 1,000 |
|
|
|
|
|
|
|
| 48 |
|
| 49 |
+
Sources are combined into single splits (no per-source or per-difficulty division).
|
| 50 |
+
`source` and `token_count` are kept as columns for optional filtering.
|
|
|
|
| 51 |
|
| 52 |
+
## Fields
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
|
| 54 |
| Field | Type | Description |
|
| 55 |
|-------|------|-------------|
|
| 56 |
+
| `id` | string | Unique example id (source-prefixed, e.g. `starvector_…`, `crawled_…`, `nn_diagram_…`) |
|
| 57 |
+
| `source` | string | `starvector` \| `crawled` \| `nn_diagram` |
|
| 58 |
+
| `token_count` | int32 | Token length of the ground-truth SVG |
|
| 59 |
+
| `image` | image | The rendered structured raster image (input; PNG, embedded) |
|
| 60 |
| `svg` | string | Ground-truth object-level SVG / XML code (target) |
|
|
|
|
|
|
|
|
|
|
| 61 |
|
| 62 |
+
## Benchmark
|
| 63 |
+
|
| 64 |
+
The 1,000-sample benchmark draws from all three sources (starvector 569, crawled 334,
|
| 65 |
+
nn_diagram 97), all with SVGs ≤ 8,192 tokens, and is stratified by ground-truth SVG length
|
| 66 |
+
into easy (≤ 2,048), medium (2,048–4,096), and hard (> 4,096) tiers (333 / 333 / 334).
|
| 67 |
+
It is the evaluation set for the paper's Tables 2–3.
|
| 68 |
+
|
| 69 |
+
> Note: the benchmark contains 1,000 rows but 997 distinct `id`s (3 rows are exact
|
| 70 |
+
> duplicates, kept to match the paper's reported count of 1,000).
|
| 71 |
|
| 72 |
## Usage
|
| 73 |
|
|
|
|
| 87 |
|
| 88 |
Predictions are scored with **image-level** metrics (render success rate, DINOv2 similarity,
|
| 89 |
LPIPS, SSIM between the rendered prediction and the input image) and a **text-level**
|
| 90 |
+
GPTScore over the SVG code. See the paper for the render-gated evaluation protocol.
|
|
|
|
| 91 |
|
| 92 |
## Licensing
|
| 93 |
|
| 94 |
StructHub aggregates multiple sources, including crawled open-domain graphics. Per-source
|
| 95 |
terms may differ; the dataset is released for **research use**. Please review and respect the
|
| 96 |
+
licenses of the original content. *(Confirm and update this field before wide distribution.)*
|
| 97 |
|
| 98 |
## Citation
|
| 99 |
|
data/benchmark-00000-of-00001.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:485fdb8370503048aeb9e70e3c22b57416eb9432cdc0ebe6fc31c515fb6b692b
|
| 3 |
+
size 45711956
|
data/train-00000-of-00009.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f1e96be6ff75e273fb6156596a7738f7888ddbbdd0c81f26a4ef7dab7d5824e9
|
| 3 |
+
size 701057030
|
data/train-00001-of-00009.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c9161f504ad350e1a69602caad06891747771105048af5356097f46d21fd7305
|
| 3 |
+
size 721033075
|
data/train-00002-of-00009.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:421ee8b2afc3bc7ff15d683f58c0ade8a285225eadf7ec6a5be530e4fe8d8694
|
| 3 |
+
size 716734988
|
data/train-00003-of-00009.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38d3869a9da8ee90b30d241184df48e084183303c49ae7a1cb1d171e461075c0
|
| 3 |
+
size 480028468
|
data/train-00004-of-00009.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:859af23d8e7ed805e7ef6f84954507f363bd8472b82be50f0c789c1812b9baa1
|
| 3 |
+
size 390119694
|
data/train-00005-of-00009.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:218ba6579240939ec39fe906e26edc49f2e716da89ca4d3841c4838276393183
|
| 3 |
+
size 398613181
|
data/train-00006-of-00009.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:48a4393ac02e21c885799e894b91486204d7aa3d0c357e1f7fad055fa66b042b
|
| 3 |
+
size 394730252
|
data/train-00007-of-00009.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:15a1db369337f15351e41d9a9ff65f88b61d3465d9af1cacf9486d90ff22714b
|
| 3 |
+
size 392596738
|
data/train-00008-of-00009.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3ead6e180d0faf62ec8b54ad63b5a0df57c85527c776fa667b2f4c817808a1ee
|
| 3 |
+
size 392249858
|