Search papers, labs, and topics across Lattice.
The paper introduces FLICKER, a hardware accelerator designed to improve the efficiency of 3D Gaussian Splatting (3DGS) rendering, particularly on resource-constrained edge devices. FLICKER employs a hardware-software co-design approach, incorporating techniques such as adaptive leader pixels, pixel-rectangle grouping, hierarchical Gaussian testing, and mixed-precision arithmetic to minimize the overhead of contribution testing. Experimental results show that FLICKER achieves significant speedup and energy efficiency gains compared to existing accelerators and edge GPUs, along with a reduction in area.
Edge devices can now render 3D Gaussian Splats nearly 20x faster and with 27x better energy efficiency thanks to FLICKER, a contribution-aware hardware accelerator.
Recently, 3D Gaussian Splatting (3DGS) has emerged as a mainstream rendering technique due to its photorealistic quality and low latency. However, processing massive numbers of non-contributing Gaussian points introduces significant computational overhead on resource-limited edge platforms, limiting its deployment in next-generation AR/VR devices. Contribution-based prior skipping alleviates this inefficiency, yet the resulting contribution-testing workload becomes prohibitive for edge execution. In this paper, we present FLICKER, a contribution-aware 3DGS accelerator based on hardware-software co-design. The proposed framework integrates adaptive leader pixels, pixel-rectangle grouping, hierarchical Gaussian testing, and a mixed-precision architecture to enable near pixel-level, contribution-driven rendering with minimal overhead. Experimental results demonstrate up to $1.5\times$ speedup, $2.6\times$ improvement in energy efficiency, and $14%$ area reduction compared with a state-of-the-art accelerator. Compared with a representative edge GPU, FLICKER achieves a $19.8\times$ speedup and $26.7\times$ higher energy efficiency.
