Summary of Optimized Deployment Of Deep Neural Networks For Visual Pose Estimation on Nano-drones, by Matteo Risso et al.
Optimized Deployment of Deep Neural Networks for Visual Pose Estimation on Nano-drones
by Matteo Risso, Francesco Daghero, Beatrice Alessandra Motetti, Daniele Jahier Pagliari, Enrico Macii, Massimo Poncino, Alessio Burrello
First submitted to arxiv on: 23 Feb 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This research proposes an automatic optimization pipeline for visual pose estimation tasks using Deep Neural Networks (DNNs) on miniaturized autonomous unmanned aerial vehicles (UAVs). The pipeline leverages two Neural Architecture Search (NAS) algorithms to explore the DNNs’ architectural space, resulting in networks that can be deployed on off-the-shelf nano-drones. The obtained networks improve the state-of-the-art by reducing inference latency by up to 3.22x at iso-error. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Miniaturized autonomous UAVs are getting popular because they’re small and can do cool things like navigate indoors or monitor people. But making them smart is hard because they’re tiny and not very powerful. This paper helps solve this problem by creating a special way to make Deep Neural Networks (DNNs) work better on these little drones. The new method uses two different ways to find the best combination of DNN parts, then tests how well it works on a small drone with a special chip. It’s faster than what’s currently possible, which is important for many real-world applications. |
Keywords
* Artificial intelligence * Inference * Optimization * Pose estimation
