basalt/scripts/basalt_verify_dataset.py

81 lines
2.7 KiB
Python
Raw Normal View History

2019-06-13 13:37:17 +02:00
#!/usr/bin/env python3
2019-06-03 13:37:53 +02:00
import sys
import math
import os
import numpy as np
dataset_path = sys.argv[1]
2019-06-13 13:37:17 +02:00
print(dataset_path)
2019-06-03 13:37:53 +02:00
timestamps = {}
exposures = {}
for sensor in ['cam0', 'cam1', 'imu0']:
data = np.loadtxt(dataset_path + '/mav0/' + sensor + '/data.csv', usecols=[0], delimiter=',', dtype=np.int64)
2019-06-13 13:37:17 +02:00
timestamps[sensor] = data
2019-06-03 13:37:53 +02:00
# check if dataset is OK...
2019-06-13 13:37:17 +02:00
for key, value in timestamps.items():
2019-06-03 13:37:53 +02:00
times = value * 1e-9
min_t = times.min()
max_t = times.max()
interval = max_t - min_t
2019-06-13 13:37:17 +02:00
diff = times[1:] - times[:-1]
print('==========================================')
print('sensor', key)
print('min timestamp', min_t)
print('max timestamp', max_t)
print('interval', interval)
print('hz', times.shape[0] / interval)
print('min time between consecutive msgs', diff.min())
print('max time between consecutive msgs', diff.max())
2019-06-03 13:37:53 +02:00
for i, d in enumerate(diff):
# Note: 0.001 is just a hacky heuristic, since we have nothing faster than 1000Hz. Should maybe be topic-specific.
if d < 0.001:
2019-06-13 13:37:17 +02:00
print("ERROR: Difference on consecutive measurements too small: {} - {} = {}".format(times[i + 1], times[i],
d) + ' in sensor ' + key)
2019-06-03 13:37:53 +02:00
# check if we have all images for timestamps
timestamp_to_topic = {}
2019-06-13 13:37:17 +02:00
for key, value in timestamps.items():
if not key.startswith('cam'):
continue
2019-06-03 13:37:53 +02:00
for v in value:
if v not in timestamp_to_topic:
timestamp_to_topic[v] = list()
timestamp_to_topic[v].append(key)
for key in timestamp_to_topic.keys():
if len(timestamp_to_topic[key]) != 2:
2019-06-13 13:37:17 +02:00
print('timestamp', key, 'has topics', timestamp_to_topic[key])
2019-06-03 13:37:53 +02:00
# check image data.
img_extensions = ['.png', '.jpg', '.webp']
2019-06-13 13:37:17 +02:00
for key, value in timestamps.items():
if not key.startswith('cam'):
continue
2019-06-03 13:37:53 +02:00
for v in value:
path = dataset_path + '/mav0/' + key + '/data/' + str(v)
img_exists = False
for e in img_extensions:
if os.path.exists(dataset_path + '/mav0/' + key + '/data/' + str(v) + e):
img_exists = True
if not img_exists:
print('No image data for ' + key + ' at timestamp ' + str(v))
exposure_file = dataset_path + '/mav0/' + key + '/exposure.csv'
if not os.path.exists(exposure_file):
print('No exposure data for ' + key)
continue
exposure_data = np.loadtxt(exposure_file, delimiter=',', dtype=np.int64)
for v in value:
idx = np.searchsorted(exposure_data[:, 0], v)
if exposure_data[idx, 0] != v:
print('No exposure data for ' + key + ' at timestamp ' + str(v))