Congratulations! Dr. Jueun Jung’s paper is accepted to IEEE Journal of Solid-State Circuits (JSSC)!

We propose ABNP, a real-time multi-modal end-to-end driving processor for efficient CNN-Transformer-Temporal acceleration under dynamic sparsity and temporal-memory overhead.

ABNP employs a Sparsity Reasoning Unit, flexible sparse-dense core orchestration with Segmented Aggregation Network,
and a Long-/Short-term Memory Unit for memory-efficient temporal attention.

ABNP achieves 10.32 frame/s and 71.3 mJ/frame, with 2.67× higher throughput and 218× lower energy than a state-of-the-art driving SoC.