Datasets:
filename stringlengths 25 124 | svg stringlengths 119 270k | text stringlengths 1 1.72k | dataset stringclasses 4
values |
|---|---|---|---|
arxiv_new__2508.03740v1_Fig2.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1350 600" style="font-family: 'Times New Roman', Times, serif; background-color: white;">
<defs>
<marker id="arrow" markerWidth="10" markerHeight="10" refX="9" refY="3" orient="auto" markerUnits="strokeWidth">
<path d="M0,0 L0,6 L9,3 z" fill="#4B7BBE" />
... | This technical diagram illustrates a hierarchical neural network architecture that processes an input $S$ of size $H \times W \times 3$ through a series of sequential stages. The flow begins with a gray-boxed initial stage containing "Patch Partition," "Linear Embedding," and "Swin Transformer Block," followed by three... | vfig_diagrams |
1404bb6d-7479-47ea-94aa-602728564e8a | <svg xmlns="http://www.w3.org/2000/svg" fill="none" height="128" stroke="#000000" stroke-width="3" viewBox="0 0 64 64" width="128"><g stroke-linecap="round"><path d="m23.71 18.56a9.37 9.37 0 1 1 9.69 14.44 2.36 2.36 0 0 0 -1.92 2.29v3.91" /><circle cx="31.85" cy="47.15" r="2.4" /><circle cx="32" cy="32" r="25.55" /></g... | A black and white icon of a question mark inside a circle. | svgx_core |
arxiv_new__2505.08327v1_KD-based_framework.svg | <svg xmlns="http://www.w3.org/2000/svg" width="900" height="750" viewBox="0 0 900 750">
<defs>
<marker id="arrow" markerWidth="10" markerHeight="10" refX="9" refY="3" orient="auto" markerUnits="strokeWidth">
<path d="M0,0 L0,6 L9,3 z" fill="black" />
</marker>
</defs>
<style>
text { font-f... | This diagram illustrates a sequential machine learning process involving two parallel model paths, labeled "Pre-trained model (teacher)" on the left and "Pre-trained model (student)" on the right. The workflow is divided into three vertical stages: "Initial task," "Knowledge distillation," and "Subsequent tasks." Solid... | vfig_diagrams |
arxiv_new__2508.18284v1_cnn_oceanic_noattention.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1200 900" style="background-color: white;">
<defs>
<marker id="arrowPurple" markerWidth="10" markerHeight="10" refX="9" refY="5" orient="auto">
<path d="M0,0 L10,5 L0,10" fill="#663399" />
</marker>
<marker id="arrowRed" markerWidth="10" mark... | This diagram illustrates a machine learning pipeline, labeled "Training/Validation Phase," which processes input shapes through a series of 2D-Convolution and Max-pooling layers. The data is then passed through a "Flatten" layer into a Feedforward Neural Network (FNN) to produce an "Output." A separate "Testing/Predict... | vfig_diagrams |
b7873464-02b2-4110-9e10-8caa8bf03189 | <svg xmlns="http://www.w3.org/2000/svg" height="128" preserveAspectRatio="xMidYMid meet" viewBox="0 0 64 64" width="128"><path d="m2 44.8c1.8 3.4 2.8 4.2 8.3 4.8 6.4.7 1.7 4.3 12.1 5.2 9.7-4 5.2-5.8 12.6-9.1 11.2-5 8.3-7.5 14.6-10.9 10.3-5.4 6.3-8.3 12.4-11.5-2-3.8-5-2.4-6.4-5.1-2.2-4.3-5.4-1.9-5.4-7.3-7.7 5-4.7 7.3-15... | A piece of bacon. | svgx_core |
arxiv_new__2511.09936v2_swap_lock.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1200 760" font-family="serif">
<defs>
<marker id="arrow-black" markerWidth="6" markerHeight="6" refX="5" refY="3" orient="auto">
<path d="M0,0 L6,3 L0,6" fill="none" stroke="black" stroke-width="1" />
</marker>
<marker id="arrow-fill-black... | This technical diagram illustrates a multi-level system architecture divided into four horizontal layers labeled "Lvl1: MS Req," "Lvl2: Swap Type," "Lvl3: MP Bitmap," and "Lvl4: Req State." The top layer features an LRU block, request queues, and a tree structure connected to a list of CPU nodes (cpu0 through cpu15). T... | vfig_diagrams |
arxiv_new__2503.07391v1_overview.svg | <svg xmlns="http://www.w3.org/2000/svg" width="900" height="500" viewBox="0 0 900 500">
<defs>
<marker id="arrowhead" markerWidth="10" markerHeight="7" refX="9" refY="3.5" orient="auto">
<polygon points="0 0, 10 3.5, 0 7" fill="black" />
</marker>
<marker id="arrowhead-start" markerWidth="... | This diagram illustrates a blockchain transaction workflow involving four main components: Client/SDK, Membership, Peer, and Orderer. The process begins with "0. Enroll" from the Membership service, followed by "1. Proposal" and "2. Submit Transaction" from the Client/SDK to the Peer. The Peer then performs "3. Relay" ... | vfig_diagrams |
ec106111-2206-4466-8023-e739a427c3e9 | <svg xmlns="http://www.w3.org/2000/svg" enable-background="new 0 0 48 48" height="128" viewBox="0 0 48 48" width="128"><path d="m7 7v34h34v-34zm32 8h-6v-6h6zm-14 0v-6h6v6zm6 2v6h-6v-6zm-8-2h-6v-6h6zm0 2v6h-6v-6zm-8 6h-6v-6h6zm0 2v6h-6v-6zm2 0h6v6h-6zm6 8v6h-6v-6zm2 0h6v6h-6zm0-2v-6h6v6zm8-6h6v6h-6zm0-2v-6h6v6zm-18-14v6... | A large light blue square contains a five by five grid of white squares. | svgx_core |
arxiv_new__2509.24378v2_Framework2.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1600 900" style="background-color: white; font-family: Arial, sans-serif;">
<defs>
<marker id="arrow-head" markerWidth="10" markerHeight="7" refX="9" refY="3.5" orient="auto">
<polygon points="0 0, 10 3.5, 0 7" fill="black" />
</marker>
<mark... | This diagram illustrates a framework for time series analysis using an LLM, featuring three types of hints: (i) Symbolic Numerical Hint, (ii) Context-Integrated Step-aligned hint, and (iii) Task-prior Hint. These hints are tokenized and fed into an LLM alongside an "LLM Input Template" and an "Arbitrary Question" to ge... | vfig_diagrams |
arxiv_new__2503.04780v1_g2t_loss_modification.svg | <svg xmlns="http://www.w3.org/2000/svg" width="1000" height="450" viewBox="0 0 1000 450">
<defs>
<marker id="arrow-black" markerWidth="10" markerHeight="10" refX="9" refY="3" orient="auto" markerUnits="strokeWidth">
<path d="M0,0 L0,6 L9,3 z" fill="#000000" />
</marker>
<marker id="arrow-red" m... | This image compares two approaches for calculating similarity: "A. Coarse-level similarity" and "B. Token-level similarity (ours)." In both diagrams, input tokens $Q_{2d}k$ and $Q_{3d}k$ are mapped to a grid of nodes representing labels like "solubility," "high," "1-C," and "[cls]." While the coarse-level approach uses... | vfig_diagrams |
paper2fig_svg__1609.04861v2-Figure4-1.svg | <svg xmlns="http://www.w3.org/2000/svg" width="900" height="550" viewBox="0 0 900 550">
<defs>
<marker id="arrowhead" markerWidth="10" markerHeight="7" refX="9" refY="3.5" orient="auto">
<polygon points="0 0, 10 3.5, 0 7" fill="#000000" />
</marker>
</defs>
<style>
text {
font-fam... | This flowchart illustrates a technical process for evaluating scene configurations using a classifier. The diagram begins with a horizontal flow from "Current Scene" to "Scene Image" and then to "Candidates' Configurations." From "Candidates' Configurations," the process moves vertically downward to "Candidates' Images... | vfig_diagrams |
2d8545ef-dd7c-4f91-bddd-c9cc0b990bcb | <svg xmlns="http://www.w3.org/2000/svg" height="128" viewBox="0 0 1024 1024" width="128"><path d="m514.85 256.87-.44-.14-4.61.07c-3.39-.54-6.04-3.53-6.04-7.05v-112.67l-194.67 194.67 194.66 194.73v-112.32a7.18 7.18 0 0 1 5.87-7.06l5.66-.14c94.04 1.02 170.25 77.87 170.25 172.15 0 94.96-77.19 172.15-172.15 172.15-61.11 0-... | The image shows a９0 degree arc with a curved line from 0 to MEDIA邀请 | svgx_core |
arxiv_new__2506.10049v2_model_0.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1000 450" font-family="Arial, Helvetica, sans-serif" font-weight="bold">
<defs>
<marker id="arrow" markerWidth="10" markerHeight="10" refX="9" refY="3" orient="auto" markerUnits="strokeWidth">
<path d="M0,0 L0,6 L9,3 z" fill="black" />
</marker>
... | This BPMN process diagram begins with a start event leading to a "Loan Requested" task, which then splits via an exclusive gateway into two parallel paths, each with a 50% probability. The top path leads to a "Manual Review" task (50 mins), while the bottom path leads to an "Auto Review" task (30 mins). Both paths conv... | vfig_diagrams |
arxiv_new__2511.17971v2_path5.svg | <svg xmlns="http://www.w3.org/2000/svg" width="1000" height="520" viewBox="0 0 1000 520">
<defs>
<marker id="arrow" markerWidth="10" markerHeight="10" refX="16" refY="3" orient="auto" markerUnits="strokeWidth">
<path d="M0,0 L0,6 L9,3 z" fill="#000000" />
</marker>
<style>
text { fo... | This diagram illustrates three different strategies for tensor contraction, labeled (a) Reconstruction contraction, (b) MAC-optimal contraction, and (c) Latency-optimal contraction. A legend at the top defines the node types: blue circles represent "Input/Output Tensor" (labeled X and Y), green circles represent "Weigh... | vfig_diagrams |
paper2fig_svg__1810.07354v1-Figure2-1.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1100 200" width="1100" height="200">
<defs>
<style>
.node-box {
fill: #FFFFFF;
stroke: #000000;
stroke-width: 2px;
}
.arrow-shape {
fill: #FFFFFF;
stroke: #000000;
stroke-width: 2px;
}
... | This diagram displays a horizontal flow chart consisting of three rounded rectangular nodes connected by two directional arrows. The first node on the left is labeled "Unreliable Environment," which points to the central node labeled "Parameter Perturbations" via an arrow containing the text "System Design." The centra... | vfig_diagrams |
arxiv_new__2511.09272v1_Conv_PE.svg | <svg xmlns="http://www.w3.org/2000/svg" width="900" height="500" viewBox="0 0 900 500" style="background-color: white;">
<defs>
<marker id="arrow" markerWidth="10" markerHeight="10" refX="9" refY="3" orient="auto" markerUnits="strokeWidth">
<path d="M0,0 L0,6 L9,3 z" fill="black" />
</marker>
... | This diagram illustrates a convolutional neural network operation where input feature maps (labeled $T_n$) are processed through a series of multipliers ($w$) and adders. The input maps are connected to a processing block containing four multipliers and three adders arranged in a tree structure. The final output consis... | vfig_diagrams |
54c3b7a6-d2f7-4e04-b21d-5ff73b194ed2 | <svg xmlns="http://www.w3.org/2000/svg" height="128" viewBox="0 0 256 256" width="128"><path d="m53.17 114.82a4 4 0 0 1 5.66-5.66l65.17 65.18v-142.34a4 4 0 0 1 8 0v142.34l65.17-65.17a4 4 0 1 1 5.66 5.66l-72 72c-.01.01-.02.01-.02.02a4.01 4.01 0 0 1 -.59.48c-.1.07-.21.11-.31.17a3.96 3.96 0 0 1 -.37.2 3.92 3.92 0 0 1 -.41... | A simple, black downward arrow icon with a single horizontal line below it, symbolizing download or descent. | svgx_core |
arxiv_new__2504.01464v1_PatchTST.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 1200 400" style="font-family: Arial, sans-serif; background-color: #FFFFFF;">
<defs>
<marker id="arrow-blue" markerWidth="10" markerHeight="10" refX="9" refY="3" orient="auto" markerUnits="strokeWidth">
<path d="M0,0 L0,6 L9,3 z" fill="#4A90E2" />
... | The image illustrates a data sequence divided into two distinct segments: "Context Length" and "Forecast Length." The "Context Length" is represented by a blue curve with discrete data points, which is further subdivided into smaller intervals labeled as "Patch Length." To the right, a pink curve extends from a moon an... | vfig_diagrams |
arxiv_new__2506.10229v1_figure1.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 850 500" font-family="Arial, Helvetica, sans-serif">
<defs>
<marker id="arrowhead" markerWidth="10" markerHeight="7" refX="9" refY="3.5" orient="auto">
<polygon points="0 0, 10 3.5, 0 7" fill="black" />
</marker>
<filter id="noise">
<feTu... | This diagram, titled "The Futures Cone" by Joseph Voros (2000), illustrates a range of future possibilities radiating from a single point labeled "Now." The cone expands over time, with nested regions representing different levels of certainty, labeled "Probable," "Preferable," "Plausible," and "Possible." The visual s... | vfig_diagrams |
548b6d27-0a18-4d84-8b3b-638f14aaf283 | <svg xmlns="http://www.w3.org/2000/svg" clip-rule="evenodd" fill-rule="evenodd" height="128" stroke-linejoin="round" stroke-miterlimit="2" viewBox="0 0 192 192" width="128"><path d="m0 0h192v192h-192z" fill="none" /><path d="m67.03 149.84s-.07-.07 0 0c4.7 4.47 12.14 4.4 16.75-.21l80.7-80.7c4.68-4.68 4.68-12.29 0-16.97l... | The image displays a simple black check mark on a white background. | svgx_core |
svg_animals_train_000292.svg | <svg xmlns="http://www.w3.org/2000/svg" width="200" height="200" viewBox="0 0 200 200">
<ellipse cx="100" cy="120" rx="45" ry="50" fill="#e8c5a0" />
<circle cx="100" cy="70" r="35" fill="#e8c5a0" />
<ellipse cx="75" cy="55" rx="21.599999999999998" ry="36.0" fill="#d4b48c" />
<ellipse cx="125" cy="55"... | Produce an svg of a tan dog - provide only the svg | animal_illustrations |
f8b50a76-78f1-423c-ab62-33333f9a1eb6 | <svg xmlns="http://www.w3.org/2000/svg" height="128" preserveAspectRatio="xMidYMid meet" viewBox="0 0 64 64" width="128"><path d="m32 2v30h-30c0 16.6 13.4 30 30 30s30-13.4 30-30-13.4-30-30-30z" fill="#ed4c5c" /><path d="m32 2c-16.6 0-30 13.4-30 30h30z" fill="#2a5f9e" /><path d="m24 20.3l5-1.3-5-1.3 3.7-3.7-5 1.3 1.3-5-... | The image is of a circular Taiwanese flag, displayed with its distinctive red field, white sun, and twelve white rays, enclosed within a blue canton. | svgx_core |
bae32788-69c3-4879-9ba2-8819e3673328 | <svg xmlns="http://www.w3.org/2000/svg" enable-background="new 0 0 128 128" viewBox="0 0 128 128"><circle cx="64" cy="64" fill="#455a64" r="60" /><path d="m34.38 68.01c17.47-10.42 44.99-8.28 59.02-1.12 6.49 3.31 9.66 6.89 17.72 11.16 6.84 3.63 7.81.7 8.23-.7 1.39-4.9 1.13-8.01 1.13-13.36.01-31.19-25.28-58.22-56.48-58.2... | A red metro train is depicted with headlights on, moving on a track. | svgx_core |
paper2fig_svg__1905.07261v1-Figure2-1.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 850 750" width="850" height="750">
<defs>
<marker id="arrow" markerWidth="10" markerHeight="10" refX="9" refY="3" orient="auto" markerUnits="strokeWidth">
<path d="M0,0 L0,6 L9,3 z" fill="#000" />
</marker>
<marker id="arrow-start" markerWidth="10... | This technical diagram illustrates a machine learning architecture divided into an "Ingredient Representation Component" on the left and a "Pairing Score Prediction Component" on the right. The left component processes "Gin" and "Tonic Water" through identical MLP blocks that share weights, resulting in two sets of out... | vfig_diagrams |
arxiv_new__2503.13486v1_train-test-split.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 800 420" width="800" height="420">
<defs>
<marker id="arrow" markerWidth="10" markerHeight="10" refX="5" refY="5" orient="auto">
<path d="M0,0 L10,5 L0,10 z" fill="black" />
</marker>
<style>
.node-box { fill: white; stroke: black;... | This diagram is a hierarchical flow chart illustrating the division of a dataset starting with a total of N=88. The top node splits into two primary branches labeled "Train" (N=58) and "Test" (N=30). Each of these branches further subdivides into two sub-categories labeled "LVO" and "NL+SM," resulting in four distinct ... | vfig_diagrams |
synthetic_svg__008847.svg | <svg xmlns="http://www.w3.org/2000/svg" width="800" height="600">
<defs>
<pattern id="fill_shape_0" patternUnits="userSpaceOnUse" width="7" height="7" patternTransform="rotate(45)">
<rect width="7" height="7" fill="#F5F5F5" />
<line x1="0" y1="0" x2="0" y2="7" stroke="#004D40" stroke-width="1" stroke-... | This flowchart illustrates a data processing workflow involving several interconnected nodes. An "Observe" block connects via a dashed line to a "Forward Optimizer" circle, which interacts with a "Process Dropout" stack and an "Ingestion Reshape" circle. The "Encoding" block connects via a dashed line to a "Cluster" ci... | vfig_shapes |
arxiv_new__2508.08435v4_fwp_primer.svg | <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 800 900" style="background-color: white;">
<defs>
<marker id="arrow-black" markerWidth="10" markerHeight="10" refX="9" refY="3" orient="auto" markerUnits="strokeWidth">
<path d="M0,0 L0,6 L9,3 z" fill="black" />
</marker>
<marker id="arrow-gr... | This image presents three architectural diagrams labeled a, b, and c. Diagram (a) shows a recurrent process where $x_t$ and $s_{t-1}$ interact with weight matrices $W^O$ and $W^R$ to produce $s_t$ and $y_t = s_t$. Diagram (b) illustrates a dynamic update mechanism where $x_t$ is transformed by $W^K, W^V,$ and $W^Q$ to ... | vfig_diagrams |
End of preview. Expand in Data Studio
TobiaSVG
TobiaSVG contains 106,524 text-SVG pairs for text-to-SVG generation and image-to-SVG vectorization. Raster images are not stored; SVGs can be rendered when examples are loaded.
Sources And Splits
| Subset | Source | Rows | License |
|---|---|---|---|
vfig_diagrams |
VFIG-Data | 59,276 | ODC-BY 1.0 |
vfig_shapes |
VFIG-Data | 5,832 | ODC-BY 1.0 |
svgx_core |
SVGX-Core-250k | 40,000 | CC BY-NC 4.0 |
animal_illustrations |
SVG Animal Illustrations | 1,416 | CC0 1.0 |
Splits contain 85,170 training, 10,599 test, and 10,755 validation rows. Identical clean SVGs are kept in the same split.
Fields
filename: original filename.svg: clean SVG source.text: text description or prompt.dataset: source subset listed above.
Processing
SVGs were normalized, parsed, and rendered; invalid rows were removed. VFIG
rows additionally required at least 40% structural primitives and at most 50
complex shapes. Existing animal prompts and SVGX Qwen captions were retained.
VFIG captions were generated from rendered images with
gemini-3.1-flash-lite. Rows were sampled with seed 42 and split by clean SVG
identity to prevent target leakage.
License
This is a mixed-license dataset. The dataset field maps each row to the source
table above. SVGX rows are CC BY-NC 4.0, so the complete dataset is for
noncommercial use. Animal illustrations are CC0 1.0. VFIG is ODC-BY 1.0, which
licenses the database but may not cover rights in every individual item; review
the ODC-BY terms and source card.
- Downloads last month
- 108