visualize_data.py

# Visualize ground-truth 6-DoF poses by projecting the spacecraft body axes
# onto dataset images.
#
# Requirements (see requirements.txt):
#   numpy
#   pandas
#   matplotlib
#   scipy
#
# Usage: python visualize_data.py [--split train|test] [--sequence RT500]
#                                 [--seed 42] [--save out.png]

import argparse
import json
import os
import random

import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
from scipy.spatial.transform import Rotation


def project(q, r, K):
    """ Projecting points to image frame to draw axes """
    # reference points in satellite frame for drawing axes
    p_axes = np.array([[0, 0, 0, 1],
                       [1, 0, 0, 1],
                       [0, 1, 0, 1],
                       [0, 0, 1, 1]])
    points_body = np.transpose(p_axes)
    # transformation to camera frame
    pose_mat = np.hstack((Rotation.from_quat(q).as_matrix(), np.expand_dims(r, 1)))
    p_cam = np.dot(pose_mat, points_body)
    # getting homogeneous coordinates
    points_camera_frame = p_cam / p_cam[2]
    # projection to image plane
    points_image_plane = K.dot(points_camera_frame)
    x, y = (points_image_plane[0], points_image_plane[1])
    return x, y


def visualize(img, q, r, K, ax=None):
    """ Visualizing image, with ground truth pose with axes projected to training image. """
    if ax is None:
        ax = plt.gca()

    ax.imshow(img)
    xa, ya = project(q, r, K)
    scale = 150
    c = np.array([[xa[0]],
                  [ya[0]]
                  ])
    p = np.array([[xa[1], xa[2], xa[3]],
                  [ya[1], ya[2], ya[3]]
                  ])
    v = p - c
    v = scale * v / np.linalg.norm(v, axis=0)
    ax.arrow(c[0, 0], c[1, 0], v[0, 0], v[1, 0], head_width=10, color='r')
    ax.arrow(c[0, 0], c[1, 0], v[0, 1], v[1, 1], head_width=10, color='g')
    ax.arrow(c[0, 0], c[1, 0], v[0, 2], v[1, 2], head_width=10, color='b')
    return


def parse_args():
    parser = argparse.ArgumentParser(description="Visualize ground-truth poses on dataset images.")
    parser.add_argument('--data-dir', default=os.path.dirname(os.path.abspath(__file__)),
                        help="Dataset root containing K.txt, train.csv and train/ (default: script directory)")
    parser.add_argument('--split', default='train', choices=['train', 'test'],
                        help="Dataset split to visualize (default: train)")
    parser.add_argument('--sequence', default=None,
                        help="Trajectory ID to visualize, e.g. RT500 (default: random)")
    parser.add_argument('--seed', type=int, default=None,
                        help="Random seed for reproducible image selection")
    parser.add_argument('--save', default=None,
                        help="Save the figure to this path instead of showing it")
    return parser.parse_args()


if __name__ == '__main__':
    args = parse_args()
    if args.seed is not None:
        random.seed(args.seed)

    split_dir = os.path.join(args.data_dir, args.split)
    sequences = sorted(d for d in os.listdir(split_dir)
                       if os.path.isdir(os.path.join(split_dir, d)))
    if not sequences:
        raise SystemExit(f"No sequence folders found in {split_dir}")

    traj_dir = args.sequence or random.choice(sequences)
    image_path = os.path.join(split_dir, traj_dir)
    if not os.path.isdir(image_path):
        raise SystemExit(f"Sequence folder not found: {image_path}")

    with open(os.path.join(args.data_dir, 'K.txt'), 'r') as file:
        K = np.array(json.load(file))

    data = pd.read_csv(os.path.join(args.data_dir, f"{args.split}.csv"))

    files = sorted(f for f in os.listdir(image_path) if f.endswith('.jpg'))
    rows = 3
    cols = 3
    picks = random.sample(files, min(rows * cols, len(files)))
    fig, axes = plt.subplots(rows, cols, figsize=(20, 20))
    for ax, image_id in zip(axes.flat, picks):
        i_data = data.loc[data['filename'] == image_id]
        if i_data.empty:
            print(f"No annotation found for {image_id}, skipping")
            continue
        r = i_data[['Tx', 'Ty', 'Tz']].to_numpy(dtype=float).squeeze()
        # Qx, Qy, Qz, Qw (scalar-last)
        q = i_data[['Qx', 'Qy', 'Qz', 'Qw']].to_numpy(dtype=float).squeeze()
        print(image_id, r, q)

        image = plt.imread(os.path.join(image_path, image_id))
        visualize(image, q, r, K, ax=ax)
    for ax in axes.flat:
        ax.axis('off')
    fig.tight_layout()
    if args.save:
        fig.savefig(args.save, bbox_inches='tight')
        print(f"Saved figure to {args.save}")
    else:
        plt.show()