日本フィジカルAI新聞

世界のフィジカルAIを、日本語で。

週刊ニュースレター購読
arXiv:2110.14491

Training Lightweight CNNs for Human-Nanodrone Proximity Interaction from Small Datasets using Background Randomization

Training Lightweight CNNs for Human-Nanodrone Proximity Interaction from Small Datasets using Background Randomization

シェア:XThreadsFacebookLINEはてブBluesky

著者: Marco Ferri, Dario Mantegazza, Elia Cereda, Nicky Zimmerman, Luca M. Gambardella, Daniele Palossi, Jérôme Guzzi, Alessandro Giusti

分類: cs.CV, cs.AI, cs.RO, stat.ML

原文アブストラクト

We consider the task of visually estimating the pose of a human from images acquired by a nearby nano-drone; in this context, we propose a data augmentation approach based on synthetic background substitution to learn a lightweight CNN model from a small real-world training set. Experimental results on data from two different labs proves that the approach improves generalization to unseen environments.