Spaces:
Runtime error
Runtime error
A newer version of the Gradio SDK is available:
6.3.0
Layout Evaluation
Quick Start
1. Full Evaluation (generate new scenes)
# Run all tests (will call AI API for config generation)
python run_evaluation.py --generate
# Run single test
python run_evaluation.py --test basic_studio_01 --generate
Note: --generate will call OpenAI API for each test to generate configuration
2. View Test Cases
python -c "from evaluation.test_cases import get_test_cases; \
cases = get_test_cases(); \
print(f'Total: {len(cases)} test cases'); \
[print(f'{i+1}. {c.id}: {c.prompt[:50]}...') for i, c in enumerate(cases)]"
Evaluation Metrics
Balanced Evaluator Scoring Components
| Metric | Weight | Description |
|---|---|---|
| Room Count Accuracy | 30% | Match between generated and GT room counts (with tolerance) |
| Room Presence | 20% | Required rooms exist |
| Adjacency | 30% | Correct adjacency relationships |
| Min Area | 10% | Meet minimum area requirements |
| Max Area | 10% | Within maximum area limits |
Score Interpretation
- 90-100%: Excellent - fully meets requirements
- 80-89%: Good - mostly meets requirements, minor issues
- 70-79%: Acceptable - meets basic requirements, room for improvement
- <70%: Insufficient - does not meet requirements
Test Cases
20 test cases covering different complexity levels and use cases:
Simple (3 cases)
basic_studio_01: Basic studio apartmentone_bedroom_apt_01: One-bedroom apartmentstudent_apartment_01: Minimal student housing
Medium (6 cases)
two_bedroom_apt_01: Two-bedroom apartmentopen_plan_loft_01: Open-plan spacefamily_home_01: Family homemaster_suite_home_01: Master suite homecompact_efficiency_01: Space-efficient compact layoutthree_bed_family_townhouse_01: Standard family townhouse
Complex (5 cases)
four_bedroom_house_01: Four-bedroom houseseparated_zones_01: Separated day/night zonesguest_suite_home_01: Guest suite homedual_master_suite_01: Dual master suitesmultigenerational_home_01: Multi-generational layout (4BR/3BA)
Special (6 cases)
home_office_layout_01: Home office layoutbalcony_apartment_01: Balcony apartmentwork_from_home_layout_01: WFH with office separationluxury_penthouse_01: Luxury penthouse with studyentertainment_home_01: Entertainment-focused designsingle_floor_accessible_01: Accessibility-friendly layout
Ground Truth Format
{
"nodes": [
{
"id": "1",
"type": "LivingRoom", # Room type
"required": True, # Is required
"count": 1, # Expected count
"min_area": 15.0, # Minimum area (sq meters)
"max_area": 30.0 # Maximum area (optional)
}
],
"edges": [
{
"from": "LivingRoom",
"to": "Kitchen",
"type": "adjacent", # Adjacency type
"required": True # Is required
}
],
"constraints": {
"adjacency": [ # Required adjacency
("LivingRoom", "Kitchen")
],
"room_counts": { # Room counts
"Bedroom": 2
},
"min_areas": { # Minimum areas
"LivingRoom": 15.0
},
"max_areas": { # Maximum areas (optional)
"LivingRoom": 30.0
}
}
}