GTC 2020: GPU-Accelerated Data Pipeline and Machine Learning on DRIVE AGX using RAPIDS
After clicking “Watch Now” you will be prompted to login or join.
Click “Watch Now” to login or join the NVIDIA Developer Program.
GPU-Accelerated Data Pipeline and Machine Learning on DRIVE AGX using RAPIDS
Andy Park, NVIDIA | Anurag Dixit, NVIDIA
We'll present the extended capability of RAPIDS on the DRIVE AGX platform by demonstrating how it enhances the in-car user experience, with examples. ML algorithms are used extensively to address various challenges in autonomous cars. They include complex, multi-stage data science pipeline, sensor data processing, modeling, and analytics to accomplish new ML-based applications. Potential applications include recommender systems through driver or vehicle personalization, visual analytics of driving data, classification of driver condition, driving scenario, and more. In many cases, application workload runs on CPUs or ECUs, leading to performance bottlenecks as data size and their computes increase. Application workload can be parallelized, causing significant speedup, using GPUs. NVIDIA developed RAPIDS to accelerate entire end-to-end data science and analytics pipelines on GPUs. In this talk, we present extended capability of RAPIDS on NVIDA DRIVE AGX platform by demonstrating how RAPIDS works.