Generates synthetic traffic light countdown reasoning tasks with four traffic lights at a crossroad (North, South, East, West) and Red/Yellow/Green cycle. The goal is to understand the 3-color traffic light cycle and predict the final state after a specified simulation time. Each light follows an independent cycle: Red (5-15s) → Yellow (3s) → Green (10s) → Yellow (3s) → Red.
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 | O-52 |
| Task | Traffic Light |
| Category | Knowledge |
| Resolution | 1024×1024 px |
| FPS | 16 fps |
| Duration | ~5-15 seconds |
| Output | PNG images + MP4 video |
# Clone the repository
git clone https://github.com/Jiaqi-Gong/Gong_VBVR_Data.git
cd Gong_VBVR_Data/O-52_traffic_light_data-generator
# Install dependencies
pip install -r requirements.txt# Generate 100 samples
python examples/generate.py --num-samples 100
# Generate with specific seed
python examples/generate.py --num-samples 100 --seed 42
# Generate without videos
python examples/generate.py --num-samples 100 --no-videos
# Custom output directory
python examples/generate.py --num-samples 100 --output data/my_output
# Generate specific task types
python examples/generate.py --num-samples 100 --types 1 2| Argument | Type | Description | Default |
|---|---|---|---|
--num-samples |
int | Number of samples to generate | Required |
--seed |
int | Random seed for reproducibility | Random |
--output |
str | Output directory | data/questions |
--no-videos |
flag | Skip video generation | False |
--types |
list[int] | Specific task types to generate (1-4) | All types |
This scene shows a crossroad with two traffic lights (North, South). Each light independently follows a 3-color cycle: Red (4s) → Yellow (4s) → Green (4s) → Yellow (4s) → Red. Currently: North light is red with 1s countdown, South light is yellow with 3s countdown. Simulate 4 seconds and show the final state of both traffic lights.
![]() |
![]() |
![]() |
| Initial Frame Traffic lights with countdown timers |
Animation Countdown decrementing to zero |
Final Frame Traffic lights after countdown reaches zero |
Understand the 3-color traffic light cycle behavior at a 4-way crossroad and predict the final state of all four lights after a specified simulation time. Each light independently cycles through states: Red → Yellow → Green → Yellow → Red.
- Traffic Lights: Four traffic lights at a crossroad (North, South, East, West)
- 3-Color Cycle: Each light independently follows: Red (5-15s random) → Yellow (3s fixed) → Green (10s fixed) → Yellow (3s fixed) → Red
- Independent Rule: Each light transitions to the next state in its cycle only when its own countdown reaches zero
- Simulation Time: Tasks specify how many seconds to simulate (e.g., "Simulate 13 seconds")
- Task Types:
- Type 1: Single light countdown (simple)
- Type 2: Two lights with countdowns (medium)
- Type 3/4: Multiple lights coordination (hard)
- Countdown Display: Shows countdown numbers that decrement over time
- State Transition: When a light's countdown reaches 0, that light transitions to the next state in its cycle
- Temporal reasoning: Tests ability to understand countdown timers and time progression
- State prediction: Requires predicting final states after countdown completion
- Independent switching: Must understand that each light switches independently
- Multiple difficulty levels: Four task types with increasing complexity
- Visual countdown: Clear numeric countdown display on traffic lights
- Smooth animation: Shows countdown decrementing frame by frame
data/questions/traffic_light_task/traffic_light_00000000/
├── first_frame.png # Initial state (traffic lights with countdown)
├── final_frame.png # Goal state (after countdown reaches zero)
├── 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 5-15 seconds long depending on countdown duration.
traffic-light countdown temporal-reasoning state-prediction time-understanding abstraction rule-inference


