Generates synthetic datasets for training and evaluating vision models on shape completion and edge detection tasks. Each sample contains geometric shapes with missing edges that must be completed by connecting visible endpoints.
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-49 |
| Task | Complete Missing Contour Segments |
| Category | Abstraction |
| Resolution | 1024×1024 px |
| FPS | 16 fps |
| Duration | ~3 seconds |
| Output | PNG images + MP4 video |
# 1. Clone the repository
git clone https://github.com/VBVR-DataFactory/G-49_complete_missing_contour_segments_data-generator.git
cd G-49_complete_missing_contour_segments_data-generator
# 2. Create and activate virtual environment
python3 -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# 3. Install dependencies
pip install --upgrade pip
pip install -r requirements.txt
pip install -e .# Generate 50 samples
python examples/generate.py --num-samples 50
# Custom output directory
python examples/generate.py --num-samples 100 --output data/my_dataset
# Reproducible generation with seed
python examples/generate.py --num-samples 50 --seed 42
# Without videos (faster)
python examples/generate.py --num-samples 50 --no-videos| Argument | Description |
|---|---|
--num-samples |
Number of tasks to generate (required) |
--output |
Output directory (default: data/questions) |
--seed |
Random seed for reproducibility |
--no-videos |
Skip video generation (images only) |
The scene shows 1 geometric shape. It has exactly one missing edge (a gap between two visible endpoints where an edge is not drawn). First locate the two endpoints of the missing edge, then draw a single straight line segment connecting them to complete the shape. Do not redraw existing edges or add any extra lines. Show the drawing process step by step.
 |
 |
 |
| Initial Frame Shapes with missing edges (gaps visible) |
Animation Missing edges drawn to complete shapes |
Final Frame All shapes completed with connected edges |
Identify and complete missing edges in geometric shapes by connecting visible endpoints with straight line segments.
- Shapes: 3 geometric figures (triangles, rectangles, etc.)
- Missing edges: Each shape has exactly one gap
- Endpoints: Two visible points where missing edge should connect
- Completion method: Draw straight line segment between endpoints
- Constraint: Do not redraw existing edges or add extra lines
- Background: White with clear shape visibility
- Goal: Complete all shapes by filling in missing edges
- Shape completion and contour closure
- Endpoint detection and connection
- Tests understanding of geometric continuity
- Multiple shapes with independent gaps
- Precise line segment drawing
- Visual reasoning for shape boundaries
data/questions/complete_missing_contour_segments_task/complete_missing_contour_segments_00000000/
├── first_frame.png # Shapes with missing edges
├── final_frame.png # Completed shapes with all edges
├── prompt.txt # Edge completion instruction
├── ground_truth.mp4 # Animation of drawing missing edges
└── question_metadata.json # Task metadata
File specifications:
- Images: 1024×1024 PNG format
- Video: MP4 format, 16 fps
- Duration: ~3 seconds
visual-reasoning shape-completion edge-detection contour-closure geometric-reasoning endpoint-connection