AI-Driven Textile Waste Classification and Circular Recovery: A Survey of Computer Vision, Recycling and Policy Literature and a Proposed Platform Architecture

The textile industry and others that use it (such as the retail sector and after-use) produce large amounts of waste at each phase of the value chain, with the majority of this waste going to landfill or incineration. The motives are not mysterious but practical: fibre composition is not usually apparent to the naked eye of the garment, sorting is still predominantly manual, and the manual methods used for the process are cumbersome and inefficient, and presently do not work with blended, contaminated or dark coloured fabrics. This paper summarizes the conclusions of 24 recent studies in automated identification and traceability, AI-based and sensor-based fibre classification, chemical- and thermo-chemical-recycling, value-added material recovery, methodology for assessing sustainability of recycling and policy for a circular economy and puts them together in a set of research gaps. We then developed a Textile Waste Intelligence Platform, an AI-based system that will help classify the fabric type and fibre composition through image analysis, rate it for recyclability and complete circularity, and proposed a recovery route with an estimated environmental value, based on the synthesis. In addition to the narrative synthesis, we extract all accuracy measures from each of the reviewed studies on classification and combine them in one common accuracy framework and quantify a gap which none of the source studies compares itself with. The accuracy drops by a mean decrease of 42.2 percentage points (range 26.6–60.9) when the same method and protocol; clean single-fibre samples change to blended, contaminated or actual post-consumer waste for all 13 accuracy decreases that were aggregated. A second comparison reveals the condition penalty is significantly greater than the 8–10 point penalty for selecting RGB imaging over spectral imaging, meaning realised accuracy is more dependent on condition of the input than the sensing technology itself. The other two patterns that repeat are disjointed classification research and disjointed sustainability scoring on incompatible system boundaries and disjointed research on recycling routes. There are no reviewed papers describing any algorithm for pricing and trading a classified batch after it has been created. Although this is a working prototype, the platform is not yet validated on actual waste streams, the validation process is described in Section 5.5 and no performance data for this platform are claimed in this paper.

Authors

Publication Details

Journal
Research Square
Published
2026-11-19
DOI
https://doi.org/10.21203/rs.3.rs-11059454/v1
Primary Topic
Waste Management and Recycling
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

AI-Driven Textile Waste Classification and Circular Recovery: A Survey of Computer Vision, Recycling and Policy Literature and a Proposed Platform Architecture

Vidhiya S, SRIVARDHAN S, Thoufeeq M, Yeshwanth Balaji B
Research Square
Waste Management and Recycling
preprint

AI-Driven Textile Waste Classification and Circular Recovery: A Survey of Computer Vision, Recycling and Policy Literature and a Proposed Platform Architecture

Vidhiya S, SRIVARDHAN S, Thoufeeq M, Yeshwanth Balaji B
preprint en

Abstract

The textile industry and others that use it (such as the retail sector and after-use) produce large amounts of waste at each phase of the value chain, with the majority of this waste going to landfill or incineration. The motives are not mysterious but practical: fibre composition is not usually apparent to the naked eye of the garment, sorting is still predominantly manual, and the manual methods used for the process are cumbersome and inefficient, and presently do not work with blended, contaminated or dark coloured fabrics. This paper summarizes the conclusions of 24 recent studies in automated identification and traceability, AI-based and sensor-based fibre classification, chemical- and thermo-chemical-recycling, value-added material recovery, methodology for assessing sustainability of recycling and policy for a circular economy and puts them together in a set of research gaps. We then developed a Textile Waste Intelligence Platform, an AI-based system that will help classify the fabric type and fibre composition through image analysis, rate it for recyclability and complete circularity, and proposed a recovery route with an estimated environmental value, based on the synthesis. In addition to the narrative synthesis, we extract all accuracy measures from each of the reviewed studies on classification and combine them in one common accuracy framework and quantify a gap which none of the source studies compares itself with. The accuracy drops by a mean decrease of 42.2 percentage points (range 26.6–60.9) when the same method and protocol; clean single-fibre samples change to blended, contaminated or actual post-consumer waste for all 13 accuracy decreases that were aggregated. A second comparison reveals the condition penalty is significantly greater than the 8–10 point penalty for selecting RGB imaging over spectral imaging, meaning realised accuracy is more dependent on condition of the input than the sensing technology itself. The other two patterns that repeat are disjointed classification research and disjointed sustainability scoring on incompatible system boundaries and disjointed research on recycling routes. There are no reviewed papers describing any algorithm for pricing and trading a classified batch after it has been created. Although this is a working prototype, the platform is not yet validated on actual waste streams, the validation process is described in Section 5.5 and no performance data for this platform are claimed in this paper.

Research Square
Waste Management and Recycling
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon