Real-Time Fully Convolutional Networks for Peer-to-Peer Nano-drone Visual Localization and LED State Estimation

会议论文
发表日期：

2024年

查看全文

Nicholas Carlotti,
Mirko Nava,
Alessandro Giusti

摘要

We design and validate a neural network for relative pose estimation and led-based communication using only visual data coming from a nano-drone’s onboard grayscale, low-resolution camera. The model is small enough to be deployed and run in real-time on an onboard hardware accelerator, achieving an inference rate of 39 FPS.