The RaDICaL dataset (Radar, Depth, IMU, RGB Camera for Learning) is an open source dataset with raw radar ADC measurements along with RGB-D (color camera and depth camera) and IMU (inertial measurement unit) measurements.
The RaDICaL dataset contains hundreds of thousands of frames with synchronized measurements providing engineers and scientists with access to feature rich data. Moreover, the radar data provided is made up of raw ADC samples unlike many other preprocessed radar datasets. By providing the raw radar measurements, users will be able to design their own processing techniques or use the methods we provide. Furthermore, the raw data retains all of its semantic information which could be very useful in deep learning applications.
Sensors
Radar: Texas Instuments IWR1443BOOST
The dataset uses an FMCW (frequency modulated continuous wave) radar. This radar has 3 transmit antennas and 4 receiving antennas. In the majority of our data, we only use 2 of the transmitting antennas on the azimuthal (horizontal) axis. Using time division multiplexing, the data can mimic a system with 1 transmitter and 8 receivers. We use custom software to collect and store the raw ADC samples.
RGB-D Camera: Intel RealSense D435i
This color/depth camera provides both high resolution color images and depth estimates of up to 10 meters. Using these images can provide ground-truths for the radar data. While calibration estimates between the radar and camera are provided, simple scenes with a radar reflector are also provided for your own calibration if you choose to do one.
Real-Time Data Capture System: Texas Instruments DCA1000EVM
In order to collect the raw ADC samples from the radar, an additional board is needed. This board processes incoming samples from the radar and sends those samples over Ethernet to the computer.