LeddarTech, a Canadian supplier of Level 1-5 ADAS and AD sensing technology, has released PixSet, which it claims is the first publicly available data set to include full-waveform data, taken from LeddarTech’s Leddar Pixell, a 3D solid-state flash lidar sensor. LeddarTech will offer these data sets free of charge for academic and research purposes.
Sensor fusion techniques are widely used to improve the performance and robustness of computer vision algorithms. Data sets such as the Leddar PixSet enable academic and engineering research teams specializing in ADAS and AD technology to use existing sets of sensor data to test and develop advanced software and to run simulations without having to assemble new sensor suites and collect their own data set.
An instrumented vehicle was used in the development of the data set, according to the company. The various scenes were recorded in high-density urban and suburban environments as well as on the highway. The data was further augmented through exposure to various weather, lighting (e.g. sunny, cloudy, rainy), and illumination (e.g. day, night, twilight) conditions. The Leddar PixSet provides information from a wide variety of situations, creating real-world data for advanced driver assistance and autonomous driving.
Data set key features include: data from an AV’s comprehensive sensor suite; full-waveform data from 3D solid-state flash lidars; 29k frames in 97 sequences, with more than 1.3M 3D boxes annotated; various environments, weather conditions and times of day; and an open-source API and data set viewer.
The Leddar PixSet was developed with the collaboration of Silicon Valley-based Deepen AI, which provided comprehensive object annotations. These new data sets provide an opportunity for 3D computer vision to go beyond lidar point clouds with a full-waveform lidar data set and are now available on the LeddarTech website.
“LeddarTech is now and has always been committed to the advancement of autonomous driving,” stated Pierre Olivier, CTO of LeddarTech. “With the release of the Leddar PixSet, we are taking one step closer to making this dream a reality. By providing these data sets free of charge to scientific and academic communities, LeddarTech is supporting and encouraging the growth and success of autonomous driving and other applications requiring lidar technology,” he continued.
“Precisely annotated data is key in the technology development of autonomous vehicles and an important step towards achieving safer roads,” added Mohammad Musa, founder & CEO at Deepen AI. “Integrating Deepen’s AI-powered annotation capabilities and LeddarTech environmental sensing platforms for autonomous vehicles, Leddar PixSet provides a diverse set of high-quality data sets to move the community forward.”