PointODE: Lightweight Point Cloud Learning with Neural Ordinary Differential Equations on Edge

TL;DR

PointODE introduces a lightweight, parameter-efficient neural network architecture based on Neural ODEs for point cloud feature extraction, optimized for embedded edge devices with FPGA acceleration, achieving high speed and energy efficiency.

Contribution

The paper proposes PointODE, a novel Neural ODE-based architecture for point cloud processing, and its FPGA accelerator, enabling efficient, accurate analysis on resource-constrained edge devices.

Findings

PointODE-Elite has 0.58M parameters and maintains competitive accuracy.

The FPGA accelerator speeds up feature extraction by 4.9x over CPU.

The approach improves energy efficiency by 3.5x compared to traditional methods.

Abstract

Embedded edge devices are often used as a computing platform to run real-world point cloud applications, but recent deep learning-based methods may not fit on such devices due to limited resources. In this paper, we aim to fill this gap by introducing PointODE, a parameter-efficient ResNet-like architecture for point cloud feature extraction based on a stack of MLP blocks with residual connections. We leverage Neural ODE (Ordinary Differential Equation), a continuous-depth version of ResNet originally developed for modeling the dynamics of continuous-time systems, to compress PointODE by reusing the same parameters across MLP blocks. The point-wise normalization is proposed for PointODE to handle the non-uniform distribution of feature points. We introduce PointODE-Elite as a lightweight version with 0.58M trainable parameters and design its dedicated accelerator for embedded FPGAs. The accelerator consists of a four-stage pipeline to parallelize the feature extraction for multiple points and stores the entire parameters on-chip to eliminate most of the off-chip data transfers. Compared to the ARM Cortex-A53 CPU, the accelerator implemented on a Xilinx ZCU104 board speeds up the feature extraction by 4.9x, leading to 3.7x faster inference and 3.5x better energy-efficiency. Despite the simple architecture, PointODE-Elite shows competitive accuracy to the state-of-the-art models on both synthetic and real-world classification datasets, greatly improving the trade-off between accuracy and inference cost.