Neural Architecture Search in operational context: a remote sensing case-study

TL;DR

This paper evaluates the effectiveness of Neural Architecture Search (NAS) in a real-world satellite imagery segmentation task, addressing hardware constraints and proposing new methodologies for operational use cases.

Contribution

It introduces a tailored NAS framework for satellite image segmentation, considering hardware constraints and providing implementation insights and novel approaches for similar operational tasks.

Findings

NAS can be adapted for satellite imagery segmentation

Hardware constraints significantly influence NAS design

Proposed methods facilitate NAS application in operational contexts

Abstract

Deep learning has become in recent years a cornerstone tool fueling key innovations in the industry, such as autonomous driving. To attain good performances, the neural network architecture used for a given application must be chosen with care. These architectures are often handcrafted and therefore prone to human biases and sub-optimal selection. Neural Architecture Search (NAS) is a framework introduced to mitigate such risks by jointly optimizing the network architectures and its weights. Albeit its novelty, it was applied on complex tasks with significant results - e.g. semantic image segmentation. In this technical paper, we aim to evaluate its ability to tackle a challenging operational task: semantic segmentation of objects of interest in satellite imagery. Designing a NAS framework is not trivial and has strong dependencies to hardware constraints. We therefore motivate our NAS approach selection and provide corresponding implementation details. We also present novel ideas to carry out other such use-case studies.