ACTIVE DEEP LEARNING FOR EFFICIENT AERIAL IMAGE LABELLING IN DISASTER

Jordan Yap; Arash Vahdat; Ferda Ofli; Patrick Meier; Greg Mori

ACTIVE DEEP LEARNING FOR EFFICIENT AERIAL IMAGE LABELLING IN DISASTER

Jordan Yap, Arash Vahdat, Ferda Ofli, Patrick Meier, Greg Mori

Qatar Computing Research Institute

Research output: Other contribution › peer-review

Abstract

Efficient analysis of aerial imagery has great potential to assist in disaster response efforts. In this paper we develop methods based on active learning to detect objects of interest. We formulate an approach based on fine-tuning deep networks for object recognition based on labels obtained from a human user. We demonstrate that this approach can effectively adapt features to recognize objects in aerial images taken after a disaster strikes.

Original language	English
Publication status	Published - 2017

Access to Document

https://scholar.google.com/citations?view_op=view_citation&hl=en&user=vgFqXEIAAAAJ&cstart=100&pagesize=100&citation_for_view=vgFqXEIAAAAJ:IUKN3-7HHlwC

Cite this

@misc{f6be5aa60a95415797152bc03e3e3fd1,

title = "ACTIVE DEEP LEARNING FOR EFFICIENT AERIAL IMAGE LABELLING IN DISASTER",

abstract = "Efficient analysis of aerial imagery has great potential to assist in disaster response efforts. In this paper we develop methods based on active learning to detect objects of interest. We formulate an approach based on fine-tuning deep networks for object recognition based on labels obtained from a human user. We demonstrate that this approach can effectively adapt features to recognize objects in aerial images taken after a disaster strikes.",

author = "Jordan Yap and Arash Vahdat and Ferda Ofli and Patrick Meier and Greg Mori",

year = "2017",

language = "English",

type = "Other",

}

TY - GEN

T1 - ACTIVE DEEP LEARNING FOR EFFICIENT AERIAL IMAGE LABELLING IN DISASTER

AU - Yap, Jordan

AU - Vahdat, Arash

AU - Ofli, Ferda

AU - Meier, Patrick

AU - Mori, Greg

PY - 2017

Y1 - 2017

N2 - Efficient analysis of aerial imagery has great potential to assist in disaster response efforts. In this paper we develop methods based on active learning to detect objects of interest. We formulate an approach based on fine-tuning deep networks for object recognition based on labels obtained from a human user. We demonstrate that this approach can effectively adapt features to recognize objects in aerial images taken after a disaster strikes.

AB - Efficient analysis of aerial imagery has great potential to assist in disaster response efforts. In this paper we develop methods based on active learning to detect objects of interest. We formulate an approach based on fine-tuning deep networks for object recognition based on labels obtained from a human user. We demonstrate that this approach can effectively adapt features to recognize objects in aerial images taken after a disaster strikes.

M3 - Other contribution

ER -

ACTIVE DEEP LEARNING FOR EFFICIENT AERIAL IMAGE LABELLING IN DISASTER

Abstract

Access to Document

Fingerprint

Cite this