tools/utils/statistics_asset_distribution.py

#!/usr/bin/env python3
"""
Statistics of asset library distribution in different datasets.
Analyzes the proportion of 3D assets from different sources.
"""

import json
import os
import sys
from pathlib import Path
from collections import defaultdict

BASE_PATH = Path("/home/v-meiszhang/backup/datas/InternScenes/Layout_info")

def extract_asset_library(uid):
    """Extract the asset library name from a model uid."""
    if uid.startswith("objaverse/"):
        return "objaverse"
    elif uid.startswith("objaverse_old/"):
        return "objaverse_old"
    elif uid.startswith("partnet_mobility"):
        return "partnet_mobility"
    elif uid.startswith("3D-FUTURE-model"):
        return "3D-FUTURE-model"
    elif uid.startswith("hssd-models"):
        return "hssd-models"
    elif uid.startswith("gen_assets"):
        return "gen_assets"
    elif uid.startswith("gr100"):
        return "gr100"
    else:
        return "unknown"

def process_dataset(dataset_name, dataset_path):
    """Process a single dataset and return statistics."""
    # Statistics containers
    total_instances = 0
    library_stats = defaultdict(lambda: {"count": 0, "scenes": set()})
    category_stats = defaultdict(lambda: defaultdict(int))
    scene_count = 0
    scenes_with_instances = 0
    
    # Handle different dataset structures
    if dataset_name == "arkitscenes":
        # ARKitScenes has Training/Validation subdirs
        search_dirs = []
        for split in ["Training", "Validation"]:
            split_path = dataset_path / split
            if split_path.exists():
                search_dirs.extend(sorted(split_path.iterdir()))
    else:
        # Other datasets have scenes directly
        search_dirs = sorted(dataset_path.iterdir())
    
    # Process all scenes
    for scene_dir in search_dirs:
        if not scene_dir.is_dir():
            continue
        
        scene_count += 1
        layout_file = scene_dir / "layout.json"
        
        if not layout_file.exists():
            continue
        
        try:
            with open(layout_file, 'r') as f:
                data = json.load(f)
            
            # layout.json is directly a list of instances
            if isinstance(data, list):
                instances = data
            else:
                instances = data.get("instances", [])
            
            if not instances:
                continue
            
            scenes_with_instances += 1
            
            for instance in instances:
                total_instances += 1
                
                uid = instance.get("model_uid", "")
                category = instance.get("category", "unknown")
                
                if not uid:
                    library = "empty_uid"
                else:
                    library = extract_asset_library(uid)
                
                library_stats[library]["count"] += 1
                library_stats[library]["scenes"].add(scene_dir.name)
                category_stats[library][category] += 1
        
        except Exception as e:
            print(f"  Error {scene_dir.name}: {e}", file=sys.stderr)
    
    return {
        "total_instances": total_instances,
        "library_stats": library_stats,
        "category_stats": category_stats,
        "scene_count": scene_count,
        "scenes_with_instances": scenes_with_instances
    }

def print_statistics(dataset_name, stats):
    """Print statistics for a dataset."""
    total_instances = stats["total_instances"]
    library_stats = stats["library_stats"]
    category_stats = stats["category_stats"]
    scene_count = stats["scene_count"]
    scenes_with_instances = stats["scenes_with_instances"]
    
    print("=" * 80)
    print(f"{dataset_name.upper()} ASSET LIBRARY STATISTICS")
    print("=" * 80)
    print()
    
    print(f"Total scenes: {scene_count}")
    print(f"Scenes with instances: {scenes_with_instances}")
    print(f"Total instances: {total_instances}")
    print()
    
    if total_instances == 0:
        print("No instances found!")
        print()
        return
    
    print("-" * 80)
    print("ASSET LIBRARY DISTRIBUTION")
    print("-" * 80)
    print()
    
    # Sort by count descending
    sorted_libs = sorted(library_stats.items(), key=lambda x: x[1]["count"], reverse=True)
    
    print(f"{'Library':<25} {'Count':<10} {'Percentage':<12} {'Scenes':<10}")
    print("-" * 80)
    
    for library, stats_data in sorted_libs:
        count = stats_data["count"]
        percentage = 100.0 * count / total_instances if total_instances > 0 else 0
        scenes_count = len(stats_data["scenes"])
        print(f"{library:<25} {count:<10} {percentage:>10.2f}%  {scenes_count:<10}")
    
    print("-" * 80)
    print()
    
    # Print top categories per library
    print("-" * 80)
    print("TOP CATEGORIES BY ASSET LIBRARY")
    print("-" * 80)
    print()
    
    for library, _ in sorted_libs[:8]:  # Top 8 libraries
        categories = category_stats[library]
        sorted_cats = sorted(categories.items(), key=lambda x: x[1], reverse=True)
        
        print(f"{library}:")
        for category, count in sorted_cats[:5]:  # Top 5 categories
            percentage = 100.0 * count / library_stats[library]["count"]
            print(f"  - {category:<30} {count:<8} ({percentage:>6.2f}%)")
        print()
    
    print("=" * 80)
    print()

def main():
    datasets = ["scannet", "arkitscenes", "3rscan", "matterport3d"]
    
    if len(sys.argv) > 1:
        # Process specific dataset if provided
        datasets = sys.argv[1:]
    
    for dataset_name in datasets:
        dataset_path = BASE_PATH / dataset_name
        
        if not dataset_path.exists():
            print(f"Warning: {dataset_path} not found, skipping")
            continue
        
        print(f"Processing {dataset_name}...")
        stats = process_dataset(dataset_name, dataset_path)
        print_statistics(dataset_name, stats)

if __name__ == "__main__":
    main()