SpikeOcc: Towards Resilient Multimodal 3D Occupancy Perception via Spiking Experts
Multimodal 3D occupancy prediction underpins safe autonomous driving through comprehensive scene understanding.
ProofPaper ↗
Key points
- In this paper, we investigate spiking neural networks (SNNs) as a means of building more resilient multimodal perception.
- To overcome this difficulty, SpikeOcc is introduced with two complementary components, a Spiking Mixture-of-Experts (SMoE) and Sensor Availability Learning (SAL).
- Specifically, SMoE organizes RGB, LiDAR, and joint fusion into specialized pathways, enabling full-width unimodal representations while retaining the original multimodal features via a bounded residual.
- SpikeOcc illustrates the promise of neuromorphic computing, pointing toward more reliable and energy-efficient autonomous driving.
Sources (1)
- [1]SpikeOcc: Towards Resilient Multimodal 3D Occupancy Perception via Spiking ExpertsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 04:28 PM
Multimodal 3D occupancy prediction underpins safe autonomous driving through comprehensive scene understanding.
In this paper, we investigate spiking neural networks (SNNs) as a means of building more resilient multimodal perception.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 11, 2026Qwen 3.8 27B Q5 vs Qwen 3.8 Next Q3_S for document analysis
- Oct 11, 2026Converting dense models into Mixture-of-Experts
- Oct 10, 2026Byte Language Models: Scaling, Emergent Abstractions, and Information Allocation
- Oct 9, 2026Beyond the Snowflakes: Disentangling Weather Effects from Confounding Meta-Factors in LiDAR Object Detection
- Oct 9, 2026Introducing Clef-omni with full multimodality, plus a faster Clef and a cheaper Clef-flash
- Oct 9, 2026Multi-Stream Attention Residuals