VRU Pose-SSD: Multiperson Pose Estimation For Automated Driving

We present a fast and efficient approach for joint person detection pose estimation optimized automated driving (AD) in urban scenarios. use multitask weight sharing architecture to jointly train estimation. This modular allows us accommodate different downstream tasks the future. By systematic large-scale experiments on Tsinghua-Daimler Urban Pose Dataset (TDUP), we obtain multiple models with varying accuracy-speed trade-offs. then quantize optimize our network deployment detailed analysis of efficacy algorithm. introduce two-stage evaluation strategy, which is more suitable AD achieve significant performance improvement comparison state-of-the-art approaches. Our model runs at 52~fps full HD images still reaches competitive 32.25~LAMR. are confident that work serves as an enabler tackle higher-level like VRU intention gesture recognition, rely stable estimates will play crucial role future systems.