Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Towards a generic film style machine learning analysis framework.
Thesis (PhD)--Stellenbosch University, 2023.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | en_ZA en_ZA |
| Published: |
Stellenbosch : Stellenbosch University
2023
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867614092240355328 |
|---|---|
| access_status_str | Open Access |
| author | Barnard, Christan |
| author2 | Nel, Gerrit Stephanus |
| author_browse | Barnard, Christan Nel, Gerrit Stephanus |
| author_facet | Nel, Gerrit Stephanus Barnard, Christan |
| author_sort | Barnard, Christan |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (PhD)--Stellenbosch University, 2023. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/126935 |
| institution | Stellenbosch University (South Africa) |
| language | en_ZA en_ZA |
| last_indexed | 2026-06-10T12:46:32.674Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2023 |
| publishDateRange | 2023 |
| publishDateSort | 2023 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/126935 Towards a generic film style machine learning analysis framework. Barnard, Christan Nel, Gerrit Stephanus Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering. Motion picture industry Machine learning Motion pictures -- Sound effects Thesis (PhD)--Stellenbosch University, 2023. ENGLISH ABSTRACT: The framework thereafter facilitates the adoption of a supervised machine learning approach and a subsequent variable importance analysis so as to identify the pre-eminent input variables to the machine learning models. This approach delivers a set of candidate salient formal stylistic parameters that may serve as accurate descriptors of the style of a specific era or director. This set of stylistic variables is further investigated by means of ancillary descriptive analyses, unsupervised machine learning analyses, and a comparison with existing film style literature pertinent to the analysis. The synthesis of quantitative and qualitative methods is facilitated by the framework so as to describe effectively and accurately the style of a specific era in film history or the work of a specific director with an appropriate level of objectivity and expository depth. The efficacy and practical workability of the proposed method are demonstrated by means of two case study applications of the film style analysis framework. The first framework instantiation comprises an analysis of film eras, with a focus on the stylistic differences between silent and early-sound films. The second case study comprises a directorial analysis of the work of Ingmar Bergman. The proposed method toward the identification of era- and director-specific salient formal stylistic parameters is demonstrated and found to exhibit significant classification and descriptive abilities. Extensive quantitative and qualitative analyses are subsequently undertaken for the purpose of validating the predictive analysis results. Expert validation of the methodology is also obtained from the originator of the statistical style analysis of films research paradigm, as well as two practicing writer-directors in the film industry. The proposed framework is therefore applied successfully toward the formal description of the stylistic deference between silent and sound films, as well as the characteristics of Ingmar Bergman's film style. AFRIKAANS OPSOMMING: Dit word tipies aanvaar onder _lmkritici, _lmontleders en _lmhistorici dat sekere _lmregisseurs 'n unieke _lmstyl vertoon. So 'n styl kan gede_nieer en ontleed word met betrekking tot formele stilistiese eienskappe soos die skaal van die skoot, kamerabewegings en redigeringspatrone wat deur die _lmmaker gebruik word. Hierdie benadering stel die ontleder in staat om insig te kry met betrekking tot die visuele styl van die _lm, asook die korpus van _lms wat vir ontleding oorweeg word. Die identi_sering van betekenisvolle formele stilistiese veranderlikes wat die werk van spesi_eke regisseurs kenmerk | sowel as algemene tendense waarneembaar in spesi_eke eras van rolprentgeskiedenis | is 'n sentrale beginsel van die analise van _lmstyl. Die gebruik van kwantitatiewe benaderings tot die identi_sering van die voorgenoemde stilistiese parameters stel die ontleder in staat om objektiewe insigte in te win met betrekking tot die styl van beide 'n individuele regisseur of 'n spesi_eke _lmera as geheel. Die ontwikkeling van berekeningsmetodes met die doel om regisseur- of era-spesi_eke betekenisvolle formele stilistiese veranderlikes te identi_seer en daarna te beskryf, is dus een van die fundamentele fasette van die statistiese stylanalise van _lms. 'n Generiese modulre raamwerk word dus in hierdie proefskrif voorgestel wat daarop gemik is om die gebruik van masjienleer en data-analise tegnieke in die analise van _lmstyl te fasiliteer en te formaliseer. Die raamwerk is ontwerp om die kon_gurasie en voorverwerking van _lmstyldata te fasiliteer, waar die data betrekking het op gemiddelde skootlengtes, verspreidings van kamerabewegings, skaal van die skoot en ander stilistiese veranderlikes wat van _lms verkryg kan word. Die raamwerk fasiliteer daarna die implimentering van 'n masjienleerbenadering en 'n daaropvolgende veranderlike belangrikheidsanalise om die beduidende insetveranderlikes tot die klassi_kasie algoritmes te identi_seer. Hierdie benadering lewer 'n stel kandidaat-betekenisvolle formele stilistiese veranderlikes wat as akkurate beskrywers van die styl van 'n spesi_eke era of regisseur kan dien. Hierdie stel stilistiese veranderlikes word verder ondersoek deur middel van bykomende beskrywende ontledings, toesiglose masjienleerontledings en 'n vergelyking met bestaande _lmstylliteratuur relevant tot die analise. Die integrasie van kwantitatiewe en kwalitatiewe metodes word deur die raamwerk gefasiliteer om die styl van 'n spesi_eke era in _lmgeskiedenis of die werk van 'n spesi_eke regisseur e_ektief en akkuraat te beskryf met die maksimum hoeveelheid objektiwiteit en uiteensettingsdiepte.Die doeltre_endheid en praktiese werkbaarheid van die voorgestelde metode word gedemonstreer in twee gevallestudietoepassings van die _lmstylanaliseraamwerk. Die eerste raamwerkinstansiasie behels 'n ontleding van die styl van _lm eras, met 'n fokus op die stilistiese verskille tussen klanklose- en vroe-klankera _lms. Die tweede gevallestudie behels 'n regisseursanalise van die werk van Ingmar Bergman. Die voorgestelde metode vir die identi_sering van era- en regisseurspesi _eke betekenisvolle formele stilistiese veranderlikes word gedemonstreer en daar word gevind dat dit noemenswaardige klassi_kasievermons toon. Deeglike en uiteenlopende kwanv titatiewe en kwalitatiewe ontledings word daarna onderneem met die doel om die masjienleer analise resultate te valideer. Deskundige validering van die metodologie word verder verkry van die oorspronklike vooraanstaander van die statistiese stylanalise van _lms navorsingsparadigma, sowel as twee praktiserende skrywer-regisseurs in die _lmbedryf. Die voorgestelde raamwerk word dus suksesvol toegepas in die formele beskrywing van die stilistiese verskille tussen klanklose- en klank_lms, asook die stylistiese kenmerke van Ingmar Bergman se _lms. Doctoral 2023-01-30T14:34:43Z 2023-05-18T06:56:32Z 2023-01-30T14:34:43Z 2023-05-18T06:56:32Z 2023-01 Thesis http://hdl.handle.net/10019.1/126935 en_ZA en_ZA Stellenbosch University xxvi, 287 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Motion picture industry Machine learning Motion pictures -- Sound effects Barnard, Christan Towards a generic film style machine learning analysis framework. |
| title | Towards a generic film style machine learning analysis framework. |
| title_full | Towards a generic film style machine learning analysis framework. |
| title_fullStr | Towards a generic film style machine learning analysis framework. |
| title_full_unstemmed | Towards a generic film style machine learning analysis framework. |
| title_short | Towards a generic film style machine learning analysis framework. |
| title_sort | towards a generic film style machine learning analysis framework |
| topic | Motion picture industry Machine learning Motion pictures -- Sound effects |
| url | http://hdl.handle.net/10019.1/126935 |
| work_keys_str_mv | AT barnardchristan towardsagenericfilmstylemachinelearninganalysisframework |