東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Landscape Quality Assessments Using ...

Levering, Alex,

Linked to FindBook

Google Book

Amazon

博客來

Landscape Quality Assessments Using Deep Learning /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Landscape Quality Assessments Using Deep Learning // Alex Levering.
Author:	Levering, Alex,
Description:	1 electronic resource (140 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-12, Section: A.
Contained By:	Dissertations Abstracts International85-12A.
Subject:	Remote sensing. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31134900
ISBN:	9798383023990

Landscape Quality Assessments Using Deep Learning /
Levering, Alex,

Landscape Quality Assessments Using Deep Learning /Alex Levering. - 1 electronic resource (140 pages)

Source: Dissertations Abstracts International, Volume: 85-12, Section: A.

Landscapes are an ever-present aspect of life, and whether for daily living or for relaxation, the qualities of landscapes influence those that use them. they are an important contributor to the health and well-being of humans. For instance, deprived living spaces can cause detrimental health effects, and landscapes with a high degree of scenic beauty are known to positively impact the health and mood of viewers. Evidently, understanding how landscapes can contribute to human well-being will have significant public health benefits. For this purpose, opinions about landscapes and their qualities need to be gathered from people. One way to do this is by showing landscape photos to volunteers, and to collect ratings from them. By then relating the contents of the landscape photos to their ratings,it can be better understood what drives the qualities of landscapes.The process of deriving inferences from landscape photos with reference scores historically required a lot of manual processing. However, recent advances in machine learning methods have enabled models to automatically extract relevant information from photos, and to use this information to perform landscape quality (LQ) ratings on new photos. As such, the deep learning models that are capable of such assessments hold tremendous potential for large-scale inventorying of LQ. However, the reasoning of these models is hard to understand, and they require an enormous amount of training examples before they are usable. Models also typically only use one source of information such as photos, while often extra information is available which can be used to train better models. These problems have thus far limited the practical applicability of deep learning models for LQ assessments. The methods brought forth in this thesis are meant to address these shortcomings, and to improve the practical utilisation of deep learning models for LQ assessments.Chapter 2 considers the prediction of live ability from aerial images for the entirety of the Netherlands. In particular, it addresses the lack of methods suitable for explaining models trained on remote sensing (RS) data. Firstly, it introduces a model which explicitly learns to relate live ability domain information (e.g. housing quality, physical environment) tolive ability ratings. This model performed better than models which learned to only predict live ability from aerial images, which demonstrates that extra task information helps to train better models. By relating the model's learned reasoning to neighbourhood typologies, it could then be understood how live ability varies for different types of neighbourhood layouts. This chapter has demonstrated that models trained on RS images can be used to extrapolate live ability surveys, and that interpretability can be part of the design process for models.

English

ISBN: 9798383023990Subjects--Topical Terms:

535394
Remote sensing.

Landscape Quality Assessments Using Deep Learning /
LDR:04172nmm a22004213i 4500 001 2430730
005 20250603111606.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798383023990
035 $a (MiAaPQD)AAI31134900
035 $a (MiAaPQD)Wageningen625396
035 $a AAI31134900
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Levering, Alex, $e author. $3 3819202
245 1 0 $a Landscape Quality Assessments Using Deep Learning / $c Alex Levering.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (140 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-12, Section: A.
500 $a Advisors: Tuia, D.; Marcos, D.
502 $b Ph.D. $c Wageningen University and Research $d 2024.
520 $a Landscapes are an ever-present aspect of life, and whether for daily living or for relaxation, the qualities of landscapes influence those that use them. they are an important contributor to the health and well-being of humans. For instance, deprived living spaces can cause detrimental health effects, and landscapes with a high degree of scenic beauty are known to positively impact the health and mood of viewers. Evidently, understanding how landscapes can contribute to human well-being will have significant public health benefits. For this purpose, opinions about landscapes and their qualities need to be gathered from people. One way to do this is by showing landscape photos to volunteers, and to collect ratings from them. By then relating the contents of the landscape photos to their ratings,it can be better understood what drives the qualities of landscapes.The process of deriving inferences from landscape photos with reference scores historically required a lot of manual processing. However, recent advances in machine learning methods have enabled models to automatically extract relevant information from photos, and to use this information to perform landscape quality (LQ) ratings on new photos. As such, the deep learning models that are capable of such assessments hold tremendous potential for large-scale inventorying of LQ. However, the reasoning of these models is hard to understand, and they require an enormous amount of training examples before they are usable. Models also typically only use one source of information such as photos, while often extra information is available which can be used to train better models. These problems have thus far limited the practical applicability of deep learning models for LQ assessments. The methods brought forth in this thesis are meant to address these shortcomings, and to improve the practical utilisation of deep learning models for LQ assessments.Chapter 2 considers the prediction of live ability from aerial images for the entirety of the Netherlands. In particular, it addresses the lack of methods suitable for explaining models trained on remote sensing (RS) data. Firstly, it introduces a model which explicitly learns to relate live ability domain information (e.g. housing quality, physical environment) tolive ability ratings. This model performed better than models which learned to only predict live ability from aerial images, which demonstrates that extra task information helps to train better models. By relating the model's learned reasoning to neighbourhood typologies, it could then be understood how live ability varies for different types of neighbourhood layouts. This chapter has demonstrated that models trained on RS images can be used to extrapolate live ability surveys, and that interpretability can be part of the design process for models.
546 $a English
590 $a School code: 2157
650 4 $a Remote sensing. $3 535394
650 4 $a Deep learning. $3 3554982
650 4 $a Educational objectives. $3 3566291
650 4 $a Neural networks. $3 677449
650 4 $a Neighborhoods. $3 993475
650 4 $a Crowdsourcing. $3 3377825
650 4 $a Semantics. $3 520060
650 4 $a Web studies. $3 2122754
650 4 $a Land use planning. $3 2122760
650 4 $a Logic. $3 529544
690 $a 0799
690 $a 0800
690 $a 0646
690 $a 0536
690 $a 0395
710 2 $a Wageningen University and Research. $e degree granting institution. $3 3815267
720 1 $a Tuia, D. $e degree supervisor.
720 1 $a Marcos, D. $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 85-12A.
790 $a 2157
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31134900