scripts/draw_bbox_overlay.py

"""
Constructive Furnace: 3D Bounding Box Visualization Tool
(空間認識AI・SLAM評価用 3Dバウンディングボックス可視化ツール)

This script reads a procedural synthetic dataset (RGB images and JSON metadata) 
and projects physically accurate 3D bounding boxes onto the 2D image plane.

Usage:
    1. Place this script in the same directory as the dataset files.
    2. Run: python draw_bbox_overlay.py --input ./ --output ./output_bbox

Dependencies:
    pip install opencv-python numpy
"""

import os
import glob
import json
import sys
import argparse

try:
    import cv2
    import numpy as np
except ImportError:
    print(" [Error] Missing required packages: pip install opencv-python numpy")
    sys.exit(1)

# ==========================================
# 設定（Settings）
# ==========================================
# キャリブレーション済みのカメラ内部パラメータ
FOCAL_LENGTH_MM = 18.0  
SENSOR_WIDTH_MM = 36.0

COLOR_MAP = {
    "Box": (0, 165, 255),       
    "Pallet": (0, 255, 0),      
    "Rack": (255, 0, 0),        
    "Unknown": (0, 0, 255),
    "DEFAULT": (0, 0, 255)      
}

# 描画用パラメータ
THICKNESS = 2
FONT = cv2.FONT_HERSHEY_SIMPLEX
FONT_SCALE = 0.5
FONT_THICKNESS = 1

# ==========================================
# 3D 座標変換・数学ロジック
# ==========================================
def get_euler_matrix(rx, ry, rz):
    cx, sx = np.cos(rx), np.sin(rx)
    cy, sy = np.cos(ry), np.sin(ry)
    cz, sz = np.cos(rz), np.sin(rz)
    
    Rx = np.array([[1, 0, 0], [0, cx, -sx], [0, sx, cx]])
    Ry = np.array([[cy, 0, sy], [0, 1, 0], [-sy, 0, cy]])
    Rz = np.array([[cz, -sz, 0], [sz, cz, 0], [0, 0, 1]])
    
    return Rz @ Ry @ Rx

def project_3d_points(local_corners, obj_pose, cam_pose, img_w, img_h):
    # オブジェクトの回転と位置
    rx, ry, rz = obj_pose["rotation_euler"]
    R_obj = get_euler_matrix(rx, ry, rz)
    T_obj = np.array(obj_pose["location"])
    
    # 8頂点を一括でワールド座標へ変換
    world_pts = (R_obj @ local_corners.T).T + T_obj
    
    # カメラの回転と位置
    R_cam = get_euler_matrix(cam_pose["rotation"]["x"], cam_pose["rotation"]["y"], cam_pose["rotation"]["z"])
    T_cam = np.array([cam_pose["location"]["x"], cam_pose["location"]["y"], cam_pose["location"]["z"]])
    
    # 一括でカメラローカル座標へ変換 (R_cam.T は逆行列と等価)
    cam_local_pts = (R_cam.T @ (world_pts - T_cam).T).T
    
    # Blender -> OpenCV の座標系変換 (Y, Zの符号を反転)
    pts_cv = np.zeros_like(cam_local_pts)
    pts_cv[:, 0] = cam_local_pts[:, 0]   
    pts_cv[:, 1] = -cam_local_pts[:, 1]  
    pts_cv[:, 2] = -cam_local_pts[:, 2]  
    
    # カメラ内部パラメータによる透視投影
    f_pixels = img_w * (FOCAL_LENGTH_MM / SENSOR_WIDTH_MM)
    cx, cy = img_w / 2.0, img_h / 2.0
    
    pts_2d = []
    for x, y, z in pts_cv:
        # カメラの背後（Z<=0.1）にある頂点は破綻を防ぐためクリッピング
        if z <= 0.1: 
            return None
            
        u = int((x / z) * f_pixels + cx)
        v = int((y / z) * f_pixels + cy)
        pts_2d.append((u, v))
        
    return pts_2d

# ==========================================
# メイン処理（汎用化バッチ）
# ==========================================
def main():
    parser = argparse.ArgumentParser(description="Draw 3D Bounding Boxes on RGB images.")
    parser.add_argument("--input", "-i", default="./", help="Directory containing RGB images and JSON metadata")
    parser.add_argument("--output", "-o", default="./output_bbox", help="Directory to save output images")
    args = parser.parse_args()

    input_dir = args.input
    output_dir = args.output

    if not os.path.exists(output_dir):
        os.makedirs(output_dir)

    search_pattern = os.path.join(input_dir, "*_RGB.png")
    rgb_files = glob.glob(search_pattern)

    if not rgb_files:
        print(f" [Warn] No RGB images found in {input_dir}. Please check the path.")
        return

    print(f" [Info] Found {len(rgb_files)} images. Starting BBox projection...")

    for rgb_path in rgb_files:
        base_name = rgb_path.replace("_RGB.png", "")
        json_path = f"{base_name}_BBox.json"
        
        if not os.path.exists(json_path):
            print(f" [Skip] JSON not found for {os.path.basename(rgb_path)}")
            continue

        with open(json_path, 'r', encoding='utf-8') as f:
            data = json.load(f)
            
        img = cv2.imread(rgb_path)
        if img is None:
            continue
            
        img_h, img_w = img.shape[:2]
        cam_pose = data.get("camera_pose", data.get("camera"))
        objects = data.get("objects", [])
        
        for obj in objects:
            # 1. クラス名の取得と色の割り当て
            class_name = obj.get("class", "DEFAULT")
            color = COLOR_MAP.get(class_name, COLOR_MAP["DEFAULT"])
            
            # 2. リスト形式 [w, d, h] として取得する
            w, d, h = obj["dimensions"]
            
            # 3. オフセット無し、純粋な中心(0,0,0)から広がる8頂点
            x_min, x_max = -w/2, w/2
            y_min, y_max = -d/2, d/2
            z_min, z_max = -h/2, h/2
            
            local_corners = np.array([
                [x_min, y_min, z_min], [x_max, y_min, z_min], [x_max, y_max, z_min], [x_min, y_max, z_min], 
                [x_min, y_min, z_max], [x_max, y_min, z_max], [x_max, y_max, z_max], [x_min, y_max, z_max]  
            ])
            
            pts_2d = project_3d_points(local_corners, obj, cam_pose, img_w, img_h)
            if pts_2d is None:
                continue 
                
            # ワイヤーフレームの描画
            edges = [(0, 1), (1, 2), (2, 3), (3, 0), (4, 5), (5, 6), (6, 7), (7, 4), (0, 4), (1, 5), (2, 6), (3, 7)]
            for start, end in edges:
                cv2.line(img, pts_2d[start], pts_2d[end], color, THICKNESS)
                
            # クラス名のテキストラベル描画
            min_x = min([p[0] for p in pts_2d])
            min_y = min([p[1] for p in pts_2d])
            label_text = f"{class_name}"
            (tw, th), baseline = cv2.getTextSize(label_text, FONT, FONT_SCALE, FONT_THICKNESS)
            
            cv2.rectangle(img, (min_x, min_y - th - 5), (min_x + tw, min_y), color, -1)
            cv2.putText(img, label_text, (min_x, min_y - 5), FONT, FONT_SCALE, (255, 255, 255), FONT_THICKNESS)

        out_filename = os.path.basename(rgb_path).replace("_RGB.png", "_BBox_Overlay.png")
        out_path = os.path.join(output_dir, out_filename)
        cv2.imwrite(out_path, img)
        print(f" [OK] Generated -> {out_filename}")

if __name__ == "__main__":
    main()