Generates synthetic tasks where shapes are displayed with some having borders and others without. The task is to identify and add thin black borders to every shape that does not already have one.
Each sample pairs a task (first frame + prompt describing what needs to happen) with its ground truth solution (final frame showing the result + video demonstrating how to achieve it). This structure enables both model evaluation and training.
| Property | Value |
|---|---|
| Task ID | G-240 |
| Task | Add Borders To Unbordered Shapes |
| Category | Transformation |
| Resolution | 1024×1024 px |
| FPS | 16 fps |
| Duration | ~6.5 seconds |
| Output | PNG images + MP4 video |
# Clone the repository
git clone https://github.com/Jiaqi-Gong/Gong_VBVR_Data.git
cd Gong_VBVR_Data/G-240_add_borders_to_unbordered_shapes_data-generator
# Install dependencies
pip install -r requirements.txt# Generate 50 samples
python examples/generate.py --num-samples 50
# Generate with specific seed
python examples/generate.py --num-samples 50 --seed 42
# Generate without videos
python examples/generate.py --num-samples 50 --no-videos
# Custom output directory
python examples/generate.py --num-samples 50 --output data/my_output| Argument | Type | Description | Default |
|---|---|---|---|
--num-samples |
int | Number of samples to generate | 50 |
--seed |
int | Random seed for reproducibility | Random |
--output |
str | Output directory | data/questions |
--no-videos |
flag | Skip video generation | False |
Several shapes are shown; some have black borders and some do not. Add a thin black border to every shape that does not already have one. Do not change anything else.
 |
 |
 |
| Initial Frame Shapes with mixed border status |
Animation Borders being added to unbordered shapes |
Final Frame All shapes now have borders |
Identify all shapes that lack borders and add a consistent thin black border to each of them, while preserving all other visual properties. This generator adds visual elements (borders) to existing shapes that lack them, demonstrating shape/element addition tasks.
- Shape Count: 3-4 shapes per image
- Shape Types: Triangle, square, circle, pentagon
- Shape Colors: Randomly selected from 60-color palette
- Shape Sizes: 80-200 pixels (radius or half-side)
- Layouts: Row, column, grid, or scatter arrangements
- Border Properties: 6-pixel black borders
- Initial State: Mixed—some shapes have borders, some don't (at least one must lack a border)
- Non-overlapping placement with 120px minimum spacing between shapes
- Margin enforcement: All shapes stay 100px away from image edges
- Adaptive layouts: Automatic size adjustment to fit shapes within bounds
- Consistent borders: All borders are 6 pixels wide and pure black
- Deduplication: Each generated sample is unique based on shape arrangement
- Smooth animation: Borders animate progressively around shapes (15 hold + 60 transition + 30 final frames)
data/questions/add_borders_to_unbordered_shapes_task/add_borders_to_unbordered_shapes_00000000/
├── first_frame.png # Initial state (shapes with mixed border status)
├── final_frame.png # Goal state (all shapes with borders)
├── prompt.txt # Task instructions
├── ground_truth.mp4 # Solution video (16 fps)
└── question_metadata.json # Task metadata
File specifications: Images are 1024×1024 PNG. Videos are MP4 at 16 fps, approximately 6.5 seconds long (105 frames total).
shape-recognition border-detection visual-completion object-properties spatial-reasoning
