EgoBench / analysis_scripts /count_stats.py
emodiary's picture
Initial complete EgoBench release
3c4b9d3 verified
Raw History Blame Contribute Delete
3.93 kB
import json
import re
import os
import sys
PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
sys.path.insert(0, PROJECT_ROOT)
from tools.warehouse import warehouse_init
from tools.household import household_init
# 1. Count tool numbers
print('=== Tool Count ===')
for scene in ['retail', 'kitchen', 'restaurant', 'warehouse', 'household']:
tool_file = 'restaurant6_tools.json' if scene == 'restaurant' else f'{scene}_tools.json'
with open(os.path.join(PROJECT_ROOT, 'tools', scene, tool_file), 'r') as f:
tools = json.load(f)
print(f'{scene}: {len(tools)}')
# 2. Count item numbers
print('\n=== Item Count ===')
# Retail
with open(os.path.join(PROJECT_ROOT, 'tools', 'retail', 'retail_init.py'), 'r') as f:
content = f.read()
products = re.findall(r'"name":\s*"([^"]+)"', content)
print(f'retail: {len(set(products))} products')
# Kitchen
with open(os.path.join(PROJECT_ROOT, 'tools', 'kitchen', 'kitchen_init.py'), 'r') as f:
content = f.read()
if '"recipes"' in content:
parts = content.split('"recipes"')
ingredients = len(re.findall(r'"name":\s*"([^"]+)"', parts[0]))
recipes = len(re.findall(r'"name":\s*"([^"]+)"', parts[1][:10000]))
print(f'kitchen: {ingredients} ingredients + {recipes} recipes')
else:
ingredients = len(re.findall(r'"name":\s*"([^"]+)"', content))
print(f'kitchen: {ingredients} ingredients')
# Restaurant
with open(os.path.join(PROJECT_ROOT, 'tools', 'restaurant', 'restaurant_init.py'), 'r') as f:
content = f.read()
if '"set_meals"' in content:
parts = content.split('"set_meals"')
dishes = len(re.findall(r'"name":\s*"([^"]+)"', parts[0]))
print(f'restaurant: {dishes} dishes')
else:
dishes = len(re.findall(r'"name":\s*"([^"]+)"', content))
print(f'restaurant: {dishes} dishes')
# Restaurant 6 (multi-restaurant variant)
with open(os.path.join(PROJECT_ROOT, 'tools', 'restaurant', 'restaurant6_init.py'), 'r') as f:
content = f.read()
if '"set_meals"' in content:
parts = content.split('"set_meals"')
dishes = len(re.findall(r'"name":\s*"([^"]+)"', parts[0]))
print(f'restaurant6: {dishes} dishes')
else:
dishes = len(re.findall(r'"name":\s*"([^"]+)"', content))
print(f'restaurant6: {dishes} dishes')
# Warehouse and household have one database snapshot per video scenario.
warehouse_count = sum(
len(getattr(warehouse_init, f'warehouse_init_data{number}')['equipment'])
for number in range(1, 26)
)
household_count = sum(
len(getattr(household_init, f'household_init_data{number}')['items'])
for number in range(1, 19)
)
print(f'warehouse: {warehouse_count} equipment records across database snapshots')
print(f'household: {household_count} item records across database snapshots')
# 3. Count scenario files
print('\n=== Scenario File Statistics ===')
scenarios_dir = os.path.join(PROJECT_ROOT, 'scenarios', 'final')
# Group by scenario
scene_stats = {}
for filename in sorted(os.listdir(scenarios_dir)):
if filename.endswith('.json'):
# Extract scenario name (remove numeric suffix)
scene_name = re.match(r'([a-z]+)', filename).group(1)
if scene_name not in scene_stats:
scene_stats[scene_name] = {'videos': set(), 'tasks': 0, 'files': 0}
filepath = os.path.join(scenarios_dir, filename)
with open(filepath, 'r') as f:
data = json.load(f)
scene_stats[scene_name]['tasks'] += len(data)
scene_stats[scene_name]['files'] += 1
for item in data:
if 'image_path' in item and item['image_path']:
scene_stats[scene_name]['videos'].add(item['image_path'])
for scene, stats in sorted(scene_stats.items()):
print(f"{scene}: videos={len(stats['videos'])}, tasks={stats['tasks']}, scenarios={stats['files']}")