utils/colmap_reader.py

"""
COLMAP Binary File Reader for ControlWave
Reads cameras.bin and images.bin files generated by VGGT/COLMAP

Based on COLMAP's official format specification and read_write_model.py
"""

import struct
import numpy as np
from pathlib import Path
from typing import Dict, List, Tuple, Optional, NamedTuple
import collections


# Data structures
CameraModel = collections.namedtuple(
    "CameraModel", ["model_id", "model_name", "num_params"]
)

Camera = collections.namedtuple(
    "Camera", ["id", "model", "width", "height", "params"]
)

BaseImage = collections.namedtuple(
    "Image", ["id", "qvec", "tvec", "camera_id", "name", "xys", "point3D_ids"]
)

Point3D = collections.namedtuple(
    "Point3D", ["id", "xyz", "rgb", "error", "image_ids", "point2D_idxs"]
)


class Image(BaseImage):
    def qvec2rotmat(self):
        return qvec2rotmat(self.qvec)


# Camera model definitions
CAMERA_MODELS = {
    0: CameraModel(model_id=0, model_name="SIMPLE_PINHOLE", num_params=3),
    1: CameraModel(model_id=1, model_name="PINHOLE", num_params=4),
    2: CameraModel(model_id=2, model_name="SIMPLE_RADIAL", num_params=4),
    3: CameraModel(model_id=3, model_name="RADIAL", num_params=5),
    4: CameraModel(model_id=4, model_name="OPENCV", num_params=8),
    5: CameraModel(model_id=5, model_name="OPENCV_FISHEYE", num_params=8),
    6: CameraModel(model_id=6, model_name="FULL_OPENCV", num_params=12),
    7: CameraModel(model_id=7, model_name="FOV", num_params=5),
    8: CameraModel(model_id=8, model_name="SIMPLE_RADIAL_FISHEYE", num_params=4),
    9: CameraModel(model_id=9, model_name="RADIAL_FISHEYE", num_params=5),
    10: CameraModel(model_id=10, model_name="THIN_PRISM_FISHEYE", num_params=12),
}

CAMERA_MODEL_IDS = CAMERA_MODELS


def read_next_bytes(fid, num_bytes, format_char_sequence, endian_character="<"):
    """Read and unpack the next bytes from a binary file.
    
    Args:
        fid: File handle
        num_bytes: Number of bytes to read
        format_char_sequence: Format string for struct.unpack
        endian_character: Endianness character ('<' for little-endian)
    
    Returns:
        Unpacked data tuple
    """
    data = fid.read(num_bytes)
    return struct.unpack(endian_character + format_char_sequence, data)


def qvec2rotmat(qvec):
    """Convert quaternion to rotation matrix.
    
    Args:
        qvec: Quaternion [w, x, y, z]
    
    Returns:
        3x3 rotation matrix
    """
    return np.array(
        [
            [
                1 - 2 * qvec[2] ** 2 - 2 * qvec[3] ** 2,
                2 * qvec[1] * qvec[2] - 2 * qvec[0] * qvec[3],
                2 * qvec[3] * qvec[1] + 2 * qvec[0] * qvec[2],
            ],
            [
                2 * qvec[1] * qvec[2] + 2 * qvec[0] * qvec[3],
                1 - 2 * qvec[1] ** 2 - 2 * qvec[3] ** 2,
                2 * qvec[2] * qvec[3] - 2 * qvec[0] * qvec[1],
            ],
            [
                2 * qvec[3] * qvec[1] - 2 * qvec[0] * qvec[2],
                2 * qvec[2] * qvec[3] + 2 * qvec[0] * qvec[1],
                1 - 2 * qvec[1] ** 2 - 2 * qvec[2] ** 2,
            ],
        ]
    )


def rotmat2qvec(R):
    """Convert rotation matrix to quaternion.
    
    Args:
        R: 3x3 rotation matrix
    
    Returns:
        Quaternion [w, x, y, z]
    """
    Rxx, Ryx, Rzx, Rxy, Ryy, Rzy, Rxz, Ryz, Rzz = R.flat
    K = (
        np.array(
            [
                [Rxx - Ryy - Rzz, 0, 0, 0],
                [Ryx + Rxy, Ryy - Rxx - Rzz, 0, 0],
                [Rzx + Rxz, Rzy + Ryz, Rzz - Rxx - Ryy, 0],
                [Ryz - Rzy, Rzx - Rxz, Rxy - Ryx, Rxx + Ryy + Rzz],
            ]
        )
        / 3.0
    )
    eigvals, eigvecs = np.linalg.eigh(K)
    qvec = eigvecs[[3, 0, 1, 2], np.argmax(eigvals)]
    if qvec[0] < 0:
        qvec *= -1
    return qvec


def read_cameras_binary(path_to_model_file):
    """Read cameras from binary file.
    
    Args:
        path_to_model_file: Path to cameras.bin file
    
    Returns:
        Dictionary mapping camera_id to Camera object
    """
    cameras = {}
    with open(path_to_model_file, "rb") as fid:
        num_cameras = read_next_bytes(fid, 8, "Q")[0]
        for _ in range(num_cameras):
            camera_properties = read_next_bytes(
                fid, num_bytes=24, format_char_sequence="iiQQ"
            )
            camera_id = camera_properties[0]
            model_id = camera_properties[1]
            model_name = CAMERA_MODEL_IDS[camera_properties[1]].model_name
            width = camera_properties[2]
            height = camera_properties[3]
            num_params = CAMERA_MODEL_IDS[model_id].num_params
            params = read_next_bytes(
                fid, num_bytes=8 * num_params, format_char_sequence="d" * num_params
            )
            cameras[camera_id] = Camera(
                id=camera_id,
                model=model_name,
                width=width,
                height=height,
                params=np.array(params),
            )
    return cameras


def read_images_binary(path_to_model_file):
    """Read images from binary file.
    
    Args:
        path_to_model_file: Path to images.bin file
    
    Returns:
        Dictionary mapping image_id to Image object
    """
    images = {}
    with open(path_to_model_file, "rb") as fid:
        num_reg_images = read_next_bytes(fid, 8, "Q")[0]
        for _ in range(num_reg_images):
            binary_image_properties = read_next_bytes(
                fid, num_bytes=64, format_char_sequence="idddddddi"
            )
            image_id = binary_image_properties[0]
            qvec = np.array(binary_image_properties[1:5])
            tvec = np.array(binary_image_properties[5:8])
            camera_id = binary_image_properties[8]
            binary_image_name = b""
            current_char = read_next_bytes(fid, 1, "c")[0]
            while current_char != b"\x00":  # Look for the ASCII 0 entry
                binary_image_name += current_char
                current_char = read_next_bytes(fid, 1, "c")[0]
            image_name = binary_image_name.decode("utf-8")
            num_points2D = read_next_bytes(fid, 8, "Q")[0]
            x_y_id_s = read_next_bytes(
                fid, num_bytes=24 * num_points2D, format_char_sequence="ddQ" * num_points2D
            )
            xys = np.column_stack(
                [x_y_id_s[0::3], x_y_id_s[1::3]]
            )
            point3D_ids = np.array(x_y_id_s[2::3])
            images[image_id] = Image(
                id=image_id,
                qvec=qvec,
                tvec=tvec,
                camera_id=camera_id,
                name=image_name,
                xys=xys,
                point3D_ids=point3D_ids,
            )
    return images


def read_points3D_binary(path_to_model_file):
    """Read 3D points from binary file.
    
    Args:
        path_to_model_file: Path to points3D.bin file
    
    Returns:
        Dictionary mapping point3D_id to Point3D object
    """
    points3D = {}
    with open(path_to_model_file, "rb") as fid:
        num_points = read_next_bytes(fid, 8, "Q")[0]
        for _ in range(num_points):
            binary_point_line_properties = read_next_bytes(
                fid, num_bytes=43, format_char_sequence="QdddBBBd"
            )
            point3D_id = binary_point_line_properties[0]
            xyz = np.array(binary_point_line_properties[1:4])
            rgb = np.array(binary_point_line_properties[4:7])
            error = binary_point_line_properties[7]
            track_length = read_next_bytes(fid, 8, "Q")[0]
            track_elems = read_next_bytes(
                fid, num_bytes=8 * track_length, format_char_sequence="ii" * track_length
            )
            image_ids = np.array(track_elems[0::2])
            point2D_idxs = np.array(track_elems[1::2])
            points3D[point3D_id] = Point3D(
                id=point3D_id,
                xyz=xyz,
                rgb=rgb,
                error=error,
                image_ids=image_ids,
                point2D_idxs=point2D_idxs,
            )
    return points3D


class COLMAPReader:
    """Helper class to read COLMAP reconstruction files."""
    
    def __init__(self, reconstruction_dir: str):
        """Initialize reader with reconstruction directory.
        
        Args:
            reconstruction_dir: Directory containing cameras.bin, images.bin, and optionally points3D.bin
        """
        self.reconstruction_dir = Path(reconstruction_dir)
        self.cameras = None
        self.images = None
        self.points3D = None
    
    def read_cameras(self) -> Dict[int, Camera]:
        """Read cameras from cameras.bin file."""
        cameras_file = self.reconstruction_dir / "cameras.bin"
        if not cameras_file.exists():
            raise FileNotFoundError(f"cameras.bin not found in {self.reconstruction_dir}")
        self.cameras = read_cameras_binary(str(cameras_file))
        return self.cameras
    
    def read_images(self) -> Dict[int, Image]:
        """Read images from images.bin file."""
        images_file = self.reconstruction_dir / "images.bin"
        if not images_file.exists():
            raise FileNotFoundError(f"images.bin not found in {self.reconstruction_dir}")
        self.images = read_images_binary(str(images_file))
        return self.images
    
    def read_points3D(self) -> Dict[int, Point3D]:
        """Read 3D points from points3D.bin file."""
        points_file = self.reconstruction_dir / "points3D.bin"
        if not points_file.exists():
            raise FileNotFoundError(f"points3D.bin not found in {self.reconstruction_dir}")
        self.points3D = read_points3D_binary(str(points_file))
        return self.points3D
    
    def read_all(self) -> Tuple[Dict[int, Camera], Dict[int, Image], Optional[Dict[int, Point3D]]]:
        """Read all available reconstruction files."""
        cameras = self.read_cameras()
        images = self.read_images()
        
        # Points3D is optional
        points3D = None
        if (self.reconstruction_dir / "points3D.bin").exists():
            points3D = self.read_points3D()
        
        return cameras, images, points3D
    
    def get_camera_poses(self) -> List[np.ndarray]:
        """Get camera poses as 4x4 transformation matrices.
        
        Returns:
            List of 4x4 numpy arrays representing camera poses (world to camera transform)
        """
        if self.images is None:
            self.read_images()
        
        poses = []
        for image_id in sorted(self.images.keys()):
            image = self.images[image_id]
            # Convert quaternion to rotation matrix
            R = qvec2rotmat(image.qvec)
            t = image.tvec
            
            # Create 4x4 transformation matrix
            pose = np.eye(4)
            pose[:3, :3] = R
            pose[:3, 3] = t
            poses.append(pose)
        
        return poses
    
    def get_intrinsics(self) -> List[np.ndarray]:
        """Get camera intrinsics as 3x3 matrices.
        
        Returns:
            List of 3x3 intrinsic matrices
        """
        if self.cameras is None:
            self.read_cameras()
        if self.images is None:
            self.read_images()
        
        intrinsics = []
        for image_id in sorted(self.images.keys()):
            image = self.images[image_id]
            camera = self.cameras[image.camera_id]
            
            # Build intrinsic matrix based on camera model
            K = np.eye(3)
            if camera.model == "SIMPLE_PINHOLE":
                # f, cx, cy
                f, cx, cy = camera.params
                K[0, 0] = f
                K[1, 1] = f
                K[0, 2] = cx
                K[1, 2] = cy
            elif camera.model == "PINHOLE":
                # fx, fy, cx, cy
                fx, fy, cx, cy = camera.params[:4]
                K[0, 0] = fx
                K[1, 1] = fy
                K[0, 2] = cx
                K[1, 2] = cy
            else:
                # For other models, use the first two parameters as fx, fy
                # and parameters 3,4 as cx, cy (if available)
                if len(camera.params) >= 2:
                    K[0, 0] = camera.params[0]
                    K[1, 1] = camera.params[1] if len(camera.params) > 1 else camera.params[0]
                if len(camera.params) >= 4:
                    K[0, 2] = camera.params[2]
                    K[1, 2] = camera.params[3]
                else:
                    # Default to image center
                    K[0, 2] = camera.width / 2
                    K[1, 2] = camera.height / 2
            
            intrinsics.append(K)
        
        return intrinsics


# Example usage
if __name__ == "__main__":
    # Example: Read COLMAP reconstruction
    reader = COLMAPReader("/path/to/reconstruction")
    
    # Read all data
    cameras, images, points3D = reader.read_all()
    
    # Get camera poses and intrinsics
    poses = reader.get_camera_poses()
    intrinsics = reader.get_intrinsics()
    
    print(f"Found {len(cameras)} cameras")
    print(f"Found {len(images)} images")
    if points3D:
        print(f"Found {len(points3D)} 3D points")
    
    # Print first camera info
    if cameras:
        first_camera = list(cameras.values())[0]
        print(f"\nFirst camera:")
        print(f"  Model: {first_camera.model}")
        print(f"  Resolution: {first_camera.width}x{first_camera.height}")
        print(f"  Parameters: {first_camera.params}")
    
    # Print first image info
    if images:
        first_image = list(images.values())[0]
        print(f"\nFirst image:")
        print(f"  Name: {first_image.name}")
        print(f"  Camera ID: {first_image.camera_id}")
        print(f"  Rotation (qvec): {first_image.qvec}")
        print(f"  Translation: {first_image.tvec}")