This thesis investigates how artificial intelligence and digital tools can support upcycling within the fashion industry, with a specific focus on design, pattern-making, and workflow feasibility. Circular fashion is often presented as a response to the industry’s environmental crisis, but the urgency of reuse strategies is sharpened by the EU’s Sustainable Products Regulation, introducing a ban on the destruction of unsold apparel from July 2026 for large companies. Upcycling, as a high-value circular strategy, therefore becomes increasingly necessary as a response to structural waste in fashion systems. At the same time, AI adoption in fashion is rising during Industry 4.0, yet it mostly remains focused on efficiency-driven optimisation, raising questions about how relevant these tools could be for upcycling practice. Building on a critical review of academic literature and industry reports, this research identifies key gaps in both circular fashion and AI-assisted design, particularly the barriers to wider upcycling adoption, and the lack of experimental testing of AI tools within constraint-driven workflows. The thesis follows a practice-based research approach, combining literature review, case study analysis, expert interviews, benchmarking of AI tools, and an experimentation phase. The experimentation tests selected publicly available generative AI and digital tools within an upcycling-led workflow. Tool performance is evaluated through criteria derived from AI-assisted design benchmarks: intent alignment, diversity of outputs, time-to-usability, volume-to-usability ratio, quality of outputs, degree of human intervention, and technical feasibility. Rather than positioning AI as a sole actor, the thesis frames it as a system that supports iteration, variation, and decision-making under constraints, while still relying on human judgment, technical knowledge, design craft, and ethical responsibility. Findings suggest that AI can enhance specific stages of upcycling workflows, especially ideation, decision support, and conscious optimisation, yet limitations remain in testing only one garment. The thesis also addresses the sustainability paradox of AI, acknowledging both its environmental cost and its circular potential. Ultimately, it contributes a design-led workflow for implementing AI into upcycling practice.
Questa tesi analizza come l’intelligenza artificiale e gli strumenti digitali possano supportare l’upcycling nell’industria della moda, con un’attenzione specifica al design, alla modellistica e alla fattibilità dei flussi di lavoro. La moda circolare viene spesso presentata come una risposta alla crisi ambientale del settore, ma l’urgenza di strategie di riuso è ulteriormente accentuata dal Regolamento europeo sui Prodotti Sostenibili, che introduce, a partire da luglio 2026, un divieto di distruzione dell’invenduto per le grandi aziende. L’upcycling, come strategia circolare ad alto valore, diventa quindi sempre più necessario come risposta allo spreco strutturale nei sistemi moda. Allo stesso tempo, l’adozione dell’IA nella moda è in crescita nell’ambito dell’Industria 4.0, ma rimane per lo più concentrata su ottimizzazioni orientate all’efficienza, sollevando interrogativi sulla reale pertinenza di questi strumenti per la pratica dell’upcycling. Basandosi su una revisione critica della letteratura accademica e dei report di settore, questa ricerca identifica lacune chiave sia nella moda circolare sia nel design assistito dall’IA, in particolare le barriere a una più ampia adozione dell’upcycling e la mancanza di sperimentazioni sull’uso di strumenti IA all’interno di flussi di lavoro guidati da vincoli. La tesi segue un approccio di ricerca basato sulla pratica, combinando revisione della letteratura, analisi di casi studio, interviste a esperti, benchmarking di strumenti IA e una fase sperimentale. La sperimentazione testa strumenti di IA generativa e strumenti digitali disponibili pubblicamente all’interno di un workflow orientato all’upcycling. Le prestazioni degli strumenti vengono valutate tramite criteri derivati da benchmark sul design assistito dall’IA: allineamento all’intento, diversità degli output, tempo necessario per ottenere un risultato utilizzabile, rapporto volume/utilizzabilità, qualità degli output, grado di intervento umano e fattibilità tecnica. Piuttosto che considerare l’IA come un attore autonomo, la tesi la inquadra come un sistema che supporta iterazione, variazione e decision-making sotto vincoli, continuando però a dipendere dal giudizio umano, dalle competenze tecniche, dall’artigianalità progettuale e dalla responsabilità etica. I risultati suggeriscono che l’IA può migliorare specifiche fasi dei workflow di upcycling, soprattutto ideazione, supporto decisionale e ottimizzazione consapevole, pur con il limite di aver testato un solo capo. La tesi affronta inoltre il paradosso di sostenibilità dell’IA, riconoscendone sia il costo ambientale sia il potenziale circolare. In definitiva, propone un workflow guidato dal design per integrare l’IA nella pratica dell’upcycling.
AI x upcycling: evaluating AI and digital tools across the upcycling design workflow
Ponebsek, Lara
2024/2025
Abstract
This thesis investigates how artificial intelligence and digital tools can support upcycling within the fashion industry, with a specific focus on design, pattern-making, and workflow feasibility. Circular fashion is often presented as a response to the industry’s environmental crisis, but the urgency of reuse strategies is sharpened by the EU’s Sustainable Products Regulation, introducing a ban on the destruction of unsold apparel from July 2026 for large companies. Upcycling, as a high-value circular strategy, therefore becomes increasingly necessary as a response to structural waste in fashion systems. At the same time, AI adoption in fashion is rising during Industry 4.0, yet it mostly remains focused on efficiency-driven optimisation, raising questions about how relevant these tools could be for upcycling practice. Building on a critical review of academic literature and industry reports, this research identifies key gaps in both circular fashion and AI-assisted design, particularly the barriers to wider upcycling adoption, and the lack of experimental testing of AI tools within constraint-driven workflows. The thesis follows a practice-based research approach, combining literature review, case study analysis, expert interviews, benchmarking of AI tools, and an experimentation phase. The experimentation tests selected publicly available generative AI and digital tools within an upcycling-led workflow. Tool performance is evaluated through criteria derived from AI-assisted design benchmarks: intent alignment, diversity of outputs, time-to-usability, volume-to-usability ratio, quality of outputs, degree of human intervention, and technical feasibility. Rather than positioning AI as a sole actor, the thesis frames it as a system that supports iteration, variation, and decision-making under constraints, while still relying on human judgment, technical knowledge, design craft, and ethical responsibility. Findings suggest that AI can enhance specific stages of upcycling workflows, especially ideation, decision support, and conscious optimisation, yet limitations remain in testing only one garment. The thesis also addresses the sustainability paradox of AI, acknowledging both its environmental cost and its circular potential. Ultimately, it contributes a design-led workflow for implementing AI into upcycling practice.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_03_Ponebsek.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Thesis PDF Document
Dimensione
16.38 MB
Formato
Adobe PDF
|
16.38 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/253359