code/generate_seenable_object_dict.py

"""
Generate seenable_obj_dict.json for all scenes.
Example:
python code/generate_seenable_object_dict.py /home/xwang378/scratch/2025/Taxonomy/Data/simulationImage/ --scene-workers 8 --camera-workers 8
"""

import os
import json
import argparse
import numpy as np
from PIL import Image
from concurrent.futures import ProcessPoolExecutor, as_completed
from multiprocessing import cpu_count


def process_camera(save_path, camera):
    """处理单个相机的数据"""
    image_dir = os.path.join(save_path, camera)
    seg_file = os.path.join(image_dir, "seg.png")
    obj_anno_file = os.path.join(image_dir, "object_annots.json")

    if not os.path.exists(seg_file) or not os.path.exists(obj_anno_file):
        return f"[Warning] Missing files in {camera}, skipped.", False
    
    if os.path.exists(os.path.join(save_path, camera, "seenable_obj_dict.json")):
        return f"[Warning] seenable_obj_dict.json already exists for {save_path.split('/')[-1]}/{camera}, skipped.", False
    
    # 读取 segmentation 图像和标注
    with open(obj_anno_file, "r") as f:
        obj_anno = json.load(f)
    obj_annos = obj_anno.get("outputs", [])

    seg = np.array(Image.open(seg_file))

    rgb_mask = seg[:, :, :3]
    
    # ⚡ 只获取唯一颜色，不计数
    unique_colors = np.unique(rgb_mask.reshape(-1, 3), axis=0)
    color_set = set(map(tuple, unique_colors))
    
    obj_dict = {
        obj_anno["object_id"]: tuple(obj_anno["color"][0:3])
        for obj_anno in obj_annos
        if tuple(obj_anno["color"][0:3]) in color_set
    }
    
    output_file = os.path.join(save_path, camera, "seenable_obj_dict.json")
    with open(output_file, "w") as f:
        json.dump(obj_dict, f, indent=4)
    
    return f"[Saved] {output_file}", True


def process_scene(image_dir, scene_name, max_workers=None):
    save_path = os.path.join(image_dir, scene_name)

    if not os.path.exists(save_path):
        print(f"[Error] Scene path not found: {save_path}")
        return

    camera_list = [x for x in os.listdir(save_path) if not x.endswith(".json")]

    print(f"Processing scene: {scene_name}")
    print(f"Found {len(camera_list)} camera folders.")

    if not camera_list:
        print(f"✅ Done processing scene: {scene_name} (no cameras found)\n")
        return

    # 并行处理所有相机
    success_count = 0
    with ProcessPoolExecutor(max_workers=max_workers) as executor:
        # 提交所有任务
        futures = {
            executor.submit(process_camera, save_path, camera): camera 
            for camera in camera_list
        }
        
        # 收集结果
        for future in as_completed(futures):
            camera = futures[future]
            try:
                message, success = future.result()
                print(message)
                if success:
                    success_count += 1
            except Exception as exc:
                print(f"[Error] {camera} generated an exception: {exc}")

    print(f"✅ Done processing scene: {scene_name} ({success_count}/{len(camera_list)} cameras processed)\n")


if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Generate seenable_obj_dict.json for all scenes.")
    parser.add_argument("image_dir", type=str, help="Directory containing the image folders")
    parser.add_argument("--scene-workers", type=int, default=None, 
                        help="Number of parallel workers for scene-level processing (default: CPU count)")
    parser.add_argument("--camera-workers", type=int, default=None,
                        help="Number of parallel workers for camera-level processing (default: CPU count)")
    args = parser.parse_args()
    
    # 获取所有场景
    batch_dir = ['zehan', 'placement', 'jiawei', 'luoxin', 'additional']
    scenes = []
    for batch in batch_dir:
        for scene in os.listdir(os.path.join(args.image_dir, batch)):
            if os.path.isdir(os.path.join(args.image_dir, batch, scene)):
                scenes.append(os.path.join(batch, scene))
    
    print(f"Found {len(scenes)} scenes to process.")
    print(f"Scene-level workers: {args.scene_workers or cpu_count()}")
    print(f"Camera-level workers: {args.camera_workers or cpu_count()}\n")
    
    # 并行处理所有场景
    with ProcessPoolExecutor(max_workers=args.scene_workers) as executor:
        futures = {
            executor.submit(process_scene, args.image_dir, scene, args.camera_workers): scene 
            for scene in scenes
        }
        
        for future in as_completed(futures):
            scene = futures[future]
            try:
                future.result()
            except Exception as exc:
                print(f"[Error] Scene {scene} generated an exception: {exc}")
    
    print("\n🎉 All scenes processed!")