LiFT: Lightweight FPGA-tailored 3D object detection based on LiDAR data

Apr 19, 2026

0

Upload by Asher

LiFT presents a lightweight, fully quantized 3D object detection algorithm for LiDAR data optimized for real-time FPGA inference. The method is designed around FPGA-specific constraints such as a 30 GMACs complexity budget, INT8 quantization, 2D cell-based processing, and limited skip connections. It combines dual-bound pillar feature extraction, reparameterizable convolutions, and a fully sparse architecture, achieving 20.73 GMACs and strong results on the NuScenes validation set (mAP 51.84%, NDS 61.01%).

Share with

LiFT: Lightweight FPGA-tailored 3D object detection based on LiDAR data

Apr 19, 2026

0

Bookmark

Share

More Options

Fullscreen

This document is user-generated content (UGC). WPS Office is not responsible for its accuracy or copyright. If you believe this content violates your rights, please use the button.