Repository navigation
Expand file tree
/
Copy pathvisualize.py
More file actions
134 lines (111 loc) · 5.89 KB
/
Copy pathvisualize.py
File metadata and controls
134 lines (111 loc) · 5.89 KB
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
# Visualize predicted vs ground-truth velocity and trajectory per sequence.
import os, sys
sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), os.pardir)))
import os
import glob
import argparse
import numpy as np
import matplotlib.pyplot as plt
from pyhocon import ConfigFactory
from datasets import SeqDataset
from utils import CPU_Unpickler, interp_xyz
from velocity_integrator import Velocity_Integrator, integrate_pos
def plot_sequence(out_path, ts, pred_vel, gt_vel, pred_traj, gt_traj):
err = pred_traj - gt_traj
pos_err = np.linalg.norm(err, axis=-1)
ate = float(np.sqrt(np.mean(pos_err ** 2)))
ate_v = float(np.sqrt(np.mean(err[:, 1] ** 2)))
ate_h = float(np.sqrt(np.mean(np.linalg.norm(err[:, [0, 2]], axis=-1) ** 2)))
final_err = float(np.linalg.norm(err[-1]))
total_dist = float(np.linalg.norm(np.diff(gt_traj, axis=0), axis=-1).sum())
drift_pct = 100.0 * final_err / total_dist if total_dist > 0 else float('nan')
fig = plt.figure(figsize=(14, 8))
gs = fig.add_gridspec(3, 2, width_ratios=[1.2, 1.0])
vel_axes = [fig.add_subplot(gs[i, 0]) for i in range(3)]
for i, name in enumerate(['x', 'y', 'z']):
vel_axes[i].plot(ts, gt_vel[:, i], label='gt', linewidth=1.0)
vel_axes[i].plot(ts, pred_vel[:, i], label='pred', linewidth=1.0, alpha=0.8)
vel_axes[i].set_ylabel(f'v{name} [m/s]')
vel_axes[i].grid(True, alpha=0.3)
vel_axes[0].legend(loc='upper right')
vel_axes[-1].set_xlabel('time [s]')
for ax in vel_axes[:-1]:
ax.tick_params(labelbottom=False)
traj_ax = fig.add_subplot(gs[:, 1])
traj_ax.plot(gt_traj[:, 0], gt_traj[:, 2], label='gt', linewidth=1.0)
traj_ax.plot(pred_traj[:, 0], pred_traj[:, 2], label='pred', linewidth=1.0, alpha=0.8)
traj_ax.scatter(gt_traj[0, 0], gt_traj[0, 2], c='g', s=40, marker='o', label='start', zorder=5)
traj_ax.scatter(gt_traj[-1, 0], gt_traj[-1, 2], c='r', s=40, marker='x', label='gt end', zorder=5)
traj_ax.scatter(pred_traj[-1, 0], pred_traj[-1, 2], c='b', s=40, marker='x', label='pred end', zorder=5)
traj_ax.set_xlabel('x [m]')
traj_ax.set_ylabel('z [m]')
traj_ax.set_aspect('equal', adjustable='datalim')
traj_ax.grid(True, alpha=0.3)
traj_ax.legend(loc='upper left')
metrics = (
f"ATE: {ate:>7.2f} m\n"
f"ATE_V: {ate_v:>7.2f} m\n"
f"ATE_H: {ate_h:>7.2f} m\n"
f"Drift: {drift_pct:>7.2f}%"
)
traj_ax.text(0.98, 0.02, metrics, transform=traj_ax.transAxes,
ha='right', va='bottom', fontsize=14, family='monospace',
bbox=dict(facecolor='white', edgecolor='#bbbbbb', boxstyle='round,pad=0.5'))
fig.tight_layout()
fig.savefig(out_path, dpi=120)
plt.close(fig)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument("--device", type=str, default="cpu", help="cuda or cpu")
parser.add_argument("--exp", type=str, default="experiments/nymeria", help="Path for AirIO netoutput")
parser.add_argument("--seqlen", type=int, default=1000, help="the length of the segment")
parser.add_argument("--dataconf", type=str, default="configs/nymeria.conf", help="the configuration of the dataset")
parser.add_argument("--savedir", type=str, default="./result/visualizations", help="Directory where the plots will be saved")
args = parser.parse_args()
print(("\n" * 3) + str(args) + ("\n" * 3))
config = ConfigFactory.parse_file(args.dataconf)
dataset_conf = config.eval
net_result_path = os.path.join(args.exp, 'net_output.pickle')
if not os.path.isfile(net_result_path):
raise FileNotFoundError(f"Unable to load the network result: {net_result_path}")
with open(net_result_path, 'rb') as handle:
inference_state_load = CPU_Unpickler(handle).load()
folder = args.savedir
os.makedirs(folder, exist_ok=True)
for data_conf in dataset_conf.data_list:
if isinstance(data_conf.data_drive, list) and len(data_conf.data_drive) == 0:
data_drive = sorted(
os.path.splitext(os.path.basename(p))[0]
for p in glob.glob(os.path.join(data_conf["data_root"], "*.pkl"))
)
else:
data_drive = list(data_conf.data_drive)
for data_name in data_drive:
print(f"\n=== {data_name} ===")
dataset = SeqDataset(
data_conf.data_root, data_name, args.device,
name=data_conf.name, duration=args.seqlen, step_size=args.seqlen,
drop_last=False, conf=dataset_conf,
)
init = dataset.get_init_value()
inference_state = inference_state_load[data_name]
gt_ts = dataset.data['time']
vel_ts = inference_state['ts']
if "coordinate" in dataset_conf.keys() and dataset_conf["coordinate"] == "body_coord":
rotation = dataset.data['gt_orientation']
net_vel = interp_xyz(gt_ts, vel_ts[:, 0], inference_state['net_vel']).to(rotation.dtype)
net_vel = rotation * net_vel
else:
net_vel = interp_xyz(gt_ts, vel_ts[:, 0], inference_state['net_vel']).to(dataset.data['velocity'].dtype)
dt = gt_ts[1:] - gt_ts[:-1]
data_inte = {"vel": net_vel, "dt": dt}
integrator_vel = Velocity_Integrator(init['pos']).to(args.device)
inf_outstate = integrate_pos(integrator_vel, data_inte, init, dataset, device=args.device)
ts = gt_ts.cpu().numpy()
pred_vel = net_vel.detach().cpu().numpy()
gt_vel = dataset.data['velocity'].cpu().numpy()
pred_traj = inf_outstate['poses'][0].detach().cpu().numpy()
gt_traj = inf_outstate['poses_gt'][1:].cpu().numpy()
out_path = os.path.join(folder, f"{data_name}.png")
plot_sequence(out_path, ts, pred_vel, gt_vel, pred_traj, gt_traj)
print(f" saved {os.path.basename(out_path)}")