ObjectMatch/pair_eval.py at main · cangumeli/ObjectMatch · 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
import argparse
import json
import os
import sys

import numpy as np
from tqdm import tqdm

from optim.common import load_matrix
from optim.solver import test
from pairwise_backend import (
    KPConfig,
    NOCPredConfig,
    ObjectIDConfig,
    OptimConfig,
    PairwiseSolver,
    PairwiseVisualizer,
)


def main(args):
    parser = argparse.ArgumentParser()
    parser.add_argument('--file', required=True)
    parser.add_argument('--test_folder', default='TestImages')
    parser.add_argument('--vis', action='store_true')
    parser.add_argument('--vis_folder', default='vis_pairs')
    parser.add_argument('--result_json', default='results.json')
    parser.add_argument('--silent', action='store_true')
    parser.add_argument('--checkpoints', default='checkpoints')
    args = parser.parse_args(args)
    print(args)

    base_dir = os.path.abspath(args.test_folder)

    samples = []
    with open(args.file) as f:
        for line in f:
            if not line.startswith('scene'):
                continue
            scene, id0, id1 = line.split()
            samples.append((scene, id0, id1))

    match_files = run_sg(base_dir, samples)

    solver = PairwiseSolver(
        KPConfig(),
        NOCPredConfig(f'{args.checkpoints}/model_sym.pth', 'configs/NOCPred.yaml'),
        ObjectIDConfig(f'{args.checkpoints}/all_5'),
        OptimConfig(verbose=not args.silent),
    )
    if args.vis:
        visualizer = PairwiseVisualizer(args.vis_folder)

    results = []
    corrects = []
    for i, ((scene, id0, id1), match_file) in tqdm(
        enumerate(zip(samples, match_files)), dynamic_ncols=True, total=len(samples)
    ):
        print(f'{i + 1} / {len(samples)}: {scene} {id0} {id1}')

        color0 = os.path.join(base_dir, scene, 'color', f'{id0}.jpg')
        color1 = os.path.join(base_dir, scene, 'color', f'{id1}.jpg')

        record0 = solver.load_record(color0)
        record1 = solver.load_record(color1)
        match_data = np.load(match_file)

        pred_pose, gn_output, extras = solver(record0, record1, match_data, ret_extra_outputs=True)

        pose0 = load_matrix(os.path.join(base_dir, scene, 'pose', f'{id0}.txt'))
        pose1 = load_matrix(os.path.join(base_dir, scene, 'pose', f'{id1}.txt'))
        gt_pose = np.linalg.inv(pose0) @ pose1
        te, ae = test(pred_pose, gt_pose)
        print(te, ae)
        print()
        corrects.append(te <= 30 and ae <= 15)

        if args.vis:
            vis_name = f'{scene}_{id0}_{id1}'
            visualizer(vis_name, record0, record1, pred_pose, gn_output, extras)
            print()

        results.append((scene, id0, id1, te, ae))

    print(np.mean(corrects))

    with open(args.result_json, 'w') as f:
        json.dump(results, f)


def run_sg(base_dir, samples):
    print('Running super-glue...')
    sg_dir = os.path.abspath('./dump_features')

    sg_paths = []
    os.makedirs(sg_dir, exist_ok=True)

    total_new = 0
    temp_txt = 'temp_sg.txt'
    with open(temp_txt, 'w') as f:
        for scene, id0, id1 in samples:
            color0 = os.path.join(base_dir, scene, 'color', f'{id0}.jpg')
            color1 = os.path.join(base_dir, scene, 'color', f'{id1}.jpg')

            pose0 = load_matrix(os.path.join(base_dir, scene, 'pose', f'{id0}.txt'))
            pose1 = load_matrix(os.path.join(base_dir, scene, 'pose', f'{id1}.txt'))
            intr = load_matrix(os.path.join(base_dir, scene, 'intrinsic_depth.txt'))[:3, :3]

            sg_path = os.path.join(sg_dir, f'{scene}_{id0}_{id1}_matches.npz')
            sg_paths.append(sg_path)
            if os.path.isfile(sg_path):
                continue
            else:
                total_new += 1
                gt_pose_sg: np.ndarray = np.linalg.inv(np.linalg.inv(pose0) @ pose1)
                line = '{} {} 0 0 {} {} {}\n'.format(
                    color0,
                    color1,
                    ' '.join(map(str, intr.ravel().tolist())),
                    ' '.join(map(str, intr.ravel().tolist())),
                    ' '.join(map(str, gt_pose_sg.ravel().tolist())),
                )
                f.write(line)

    print(f'{len(samples) - total_new} / {len(samples)} keypoints are recovered from {sg_dir}')
    if total_new:
        os.system(
            f'{sys.executable} -u'
            + ' ../SuperGluePretrainedNetwork/match_pairs_scannet.py'
            + f' --input_dir "/" --input_pairs {temp_txt} --output_dir {sg_dir}'
        )
    os.unlink(temp_txt)
    print(len(sg_paths))
    return sg_paths


if __name__ == '__main__':
    main(sys.argv[1:])