XLD_code/convert_XLD_transforms.py at main · lifuguan/XLD_code · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
import enum
import os
import numpy as np
import json
import os
import numpy as np
from scipy.spatial.transform import Rotation as R

ORIGINAL_SIZE = [[1280, 1920], [1280, 1920], [1280, 1920], [884, 1920], [884, 1920]]


def pose_unreal2opencv(c2w_mat):

    translation = c2w_mat[:3, 3]
    rot = R.from_matrix(c2w_mat[:3, :3])
    rot_vec = rot.as_rotvec()

    rot_vec_new = rot_vec[[1, 2, 0]]
    rot_vec_new[0] *= -1
    rot_vec_new[2] *= -1

    rot = R.from_rotvec(rot_vec_new)


    translation_new = translation[[1, 2, 0]]
    translation_new[1] *= -1

    c2w_mat = np.eye(4)
    c2w_mat[:3, :3] = rot.as_matrix()
    c2w_mat[:3, 3] = translation_new

    rot = np.eye(4)
    rot[1,1]=-1
    rot[2, 2] = -1
    c2w_mat = c2w_mat @ rot
    return c2w_mat

def load_calibrations(scene_path, data_set, start_timestep, end_timestep, num_cams=1, camera_list=[0]):
    """
    Load the camera intrinsics, extrinsics, timestamps, etc.
    Compute the camera-to-world matrices, ego-to-world matrices, etc.
    """
    # to store per-camera intrinsics and extrinsics
    _intrinsics = []
    cam_to_egos = []
    for i in range(num_cams):
        # load camera intrinsics
        # 1d Array of [f_u, f_v, c_u, c_v, k{1, 2}, p{1, 2}, k{3}].
        # ====!! we did not use distortion parameters for simplicity !!====
        # to be improved!!
        intrinsic = np.loadtxt(
            os.path.join(scene_path, "intrinsics", f"{i}.txt")
        )
        fx, fy, cx, cy = intrinsic[0], intrinsic[1], intrinsic[2], intrinsic[3]
        intrinsic = np.array([[fx, 0, cx,0], [0, fy, cy,0], [0, 0, 1, 0], [0, 0, 0, 1]])
        _intrinsics.append([fx, fy, cx, cy])

        # load camera extrinsics
        cam_to_ego = np.loadtxt(
            os.path.join(scene_path, "extrinsics", f"{i}.txt")
        )
        # because we use opencv coordinate system to generate camera rays,
        # we need a transformation matrix to covnert rays from opencv coordinate
        # system to waymo coordinate system.
        # opencv coordinate system: x right, y down, z front
        # waymo coordinate system: x front, y left, z up
        cam_to_egos.append(cam_to_ego)

    # compute per-image poses and intrinsics
    cam_to_worlds= []
    intrinsics, cam_ids = [], []
    # ===! for waymo, we simplify timestamps as the time indices
    timestamps, timesteps = [], []
    prefix = "train" if "train" in data_set else "eval"
    for t in range(start_timestep, end_timestep):
        with open(os.path.join(scene_path, data_set, f"{prefix}_camera_extrinsics_{t:06d}.json"), 'r') as file:
            ego_to_world_current = np.array(json.load(file)['transform_matrix'])
        for cam_id in camera_list:
            cam_ids.append(cam_id)
            # transformation:
            #   (opencv_cam -> waymo_cam -> waymo_ego_vehicle) -> current_world

            rot = np.array([
                [1, 0, 0, 0],
                [0, 1, 0, 0],
                [0, 0, -1, 0],
                [0, 0, 0, 1],])
            cam_to_ego = cam_to_egos[cam_id]
            cam2world = pose_unreal2opencv(ego_to_world_current @ cam_to_ego)
            # cam2world = ego_to_world_current @ cam_to_ego
            cam_to_worlds.append(cam2world)

            intrinsics.append(_intrinsics[cam_id])
            # ===! we use time indices as the timestamp for waymo dataset for simplicity
            # ===! we can use the actual timestamps if needed
            # to be improved
            timestamps.append(t - start_timestep)
            timesteps.append(t - start_timestep)

    return intrinsics, cam_to_worlds, cam_ids

def generate_json_file(scene_path, num_cams, camera_list, offset_meters = 0):
    json_content = {"camera_model": "OPENCV", "frames": []}
    intrinsics, cam_to_worlds, cam_ids = load_calibrations(scene_path, "train_pic", 5, 155, num_cams=num_cams, camera_list=camera_list)
    for i, (intrinsic, c2w, cam_id) in enumerate(zip(intrinsics, cam_to_worlds, cam_ids)):
        frame = {
            "file_path": f"../{scene_path}/train_pic/train_camera{cam_id}_{(i // num_cams + 5):05d}.png",
            "transform_matrix": c2w.tolist(),
            "fl_x": intrinsic[0], # focal length x
            "fl_y": intrinsic[1], # focal length y
            "cx": intrinsic[2], # principal point x
            "cy": intrinsic[3], # principal point y
            "h": ORIGINAL_SIZE[cam_id][0],
            "w": ORIGINAL_SIZE[cam_id][1],
        }
        json_content['frames'].append(frame)

    test_num_cams, test_camera_list =1, [0]
    intrinsics, cam_to_worlds, cam_ids = load_calibrations(scene_path, f"test_pic/offset_left_{offset_meters}m", 2, 30, num_cams=1, camera_list=[0])
    for i, (intrinsic, c2w, cam_id) in enumerate(zip(intrinsics, cam_to_worlds, cam_ids)):
        if i // test_num_cams == 30:
            break
        frame = {
            "file_path": f"../{scene_path}/test_pic/offset_left_{offset_meters}m/eval_camera{cam_id}_{(i // test_num_cams + 2):05d}.png",
            "transform_matrix": c2w.tolist(),
            "fl_x": intrinsic[0], # focal length x
            "fl_y": intrinsic[1], # focal length y
            "cx": intrinsic[2], # principal point x
            "cy": intrinsic[3], # principal point y
            "h": ORIGINAL_SIZE[cam_id][0],
            "w": ORIGINAL_SIZE[cam_id][1],
        }
        json_content['frames'].append(frame)

    json_path = scene_path.split("/")[-1]+f"_cam{num_cams}"
    os.makedirs(json_path, exist_ok=True)
    with open(os.path.join(json_path, "transforms.json"),"w") as f:
        f.write(json.dumps(json_content,indent=4))

generate_json_file(scene_path = 'data/carla_pic_0603_Town01', num_cams=1, camera_list=[0], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town02', num_cams=1, camera_list=[0], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town03', num_cams=1, camera_list=[0], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town04', num_cams=1, camera_list=[0], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town05', num_cams=1, camera_list=[0], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town10', num_cams=1, camera_list=[0], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town01', num_cams=3, camera_list=[1,0,2], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town02', num_cams=3, camera_list=[1,0,2], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town03', num_cams=3, camera_list=[1,0,2], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town04', num_cams=3, camera_list=[1,0,2], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town05', num_cams=3, camera_list=[1,0,2], offset_meters=1)
generate_json_file(scene_path = 'data/carla_pic_0603_Town10', num_cams=3, camera_list=[1,0,2], offset_meters=1)