NIMBUS: A Hybrid Cloud-Crowd Realtime Architecture for Visual Learning in Interactive Domains

TL;DR

This paper introduces NIMBUS, a hybrid cloud-crowd architecture for real-time visual learning in interactive domains, demonstrating improved quality at the expense of response time through empirical experiments.

Contribution

It presents a novel hybrid architecture that balances quality and response time in cloud-crowd visual learning systems, supported by empirical validation.

Findings

Improved quality of visual learning responses

Trade-off between response time and quality demonstrated

Empirical validation using Amazon Mechanical Turk

Abstract

Robotic architectures that incorporate cloud-based resources are just now gaining popularity. However, researchers have very few investigations into their capabilities to support claims of their feasibility. We propose a novel method to exchange quality for speed of response. Further, we back this assertion with empirical findings from experiments performed with Amazon Mechanical Turk and find that our method improves quality in exchange for response time in our cognitive architecture.