Latency optimized architectures for a real-time inference pipeline for control tasks | IEEE Conference Publication | IEEE Xplore