Training Lightweight CNNs for Human-Nanodrone Proximity Interaction from Small Datasets using Background Randomization
arXiv:2110.14491
Abstract
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.
References in corpus (6)
- Unsupervised Data Augmentation for Consistency Training
- Data Augmentation using Random Image Cropping and Patching for Deep CNNs
- Smart Augmentation - Learning an Optimal Data Augmentation Strategy
- Deep Drone Racing: From Simulation to Reality with Domain Randomization
- A 64mW DNN-based Visual Navigation Engine for Autonomous Nano-Drones
- Active Domain Randomization