The RaDICaL Dataset: A synchronized and calibrated low-level Radar, RGB-D and IMU dataset.
This is pre-alpha research quality code, and is being actively developed on. Anything may change at anytime. Please check back here often.
Bug reports are very much appreciated.
- Radar config reader
- Read from aligned H5 dataset
- Read from raw bags (to produce aligned/unaligned H5 datasets)
- Polar to Cartesian Projection
- Camera/Radar coordinate transforms
python -m pip install git+https://github.com/moodoki/radical_sdk.git
Our dataset is distrbuted under CC BY 4.0 license. The dataset is currently under review and will be made available soon. DOI: 10.13012/B2IDB-3289560_V1.
Download the dataset at our project page.
A small sample (50 frames) to try things our can be found here. [md5sum: b195ff422cc4c979eeb81623899050cb]
Reading and displaying depth and RGB is easy
import matplotlib.pyplot as plt
import numpy as np
plt.rcParams['figure.figsize'] = [10, 20]
from radicalsdk.h5dataset import H5DatasetLoader
data = H5DatasetLoader('../samples/indoor_sample_50.h5')
frame_idx = 1
plt.figure()
plt.imshow(data['rgb'][frame_idx][..., ::-1])
plt.title(f'RGB frame {frame_idx}@{data["rgb_timestamp"][frame_idx]}')
plt.show()
plt.figure()
plt.imshow(data['depth'][frame_idx])
plt.title(f'Depth frame {frame_idx}@{data["depth_timestamp"][frame_idx]}')
plt.show()
The radar data is a 3D array arranged slow_time, antenna, fast_time.
RadarFrame encapsulates the necessary processing and saves computation on subsequent calls
as steps can be very expensive.
from radicalsdk.radar.config_v1 import read_radar_params
from radicalsdk.radar.v1 import RadarFrame
# Read config and configure RadarFrame object
radar_config = read_radar_params('../samples/indoor_human_rcs.cfg')
rf = RadarFrame(radar_config)
plt.figure()
plt.imshow(np.log(np.abs(rf.compute_range_azimuth(data['radar'][1]))))
plt.show()
from radicalsdk.geometry import PolarToCartesianWarp
p2c = PolarToCartesianWarp()
cartesian_radar = p2c(np.abs(rf.range_azimuth_capon)[np.newaxis, ..., np.newaxis])
plt.figure()
with np.errstate(divide='ignore'):
plt.imshow(np.log(cartesian_radar[0, ...]))
plt.show()
