BRENNEN

Industries / Food, farming and forestry / Agriculture and farming / Cotton ginning

Grading fiber

Grading measures a sample from every bale for color, trash, fiber length, strength and fineness, and those numbers set the bale's price.

7

Milestones achieved, from 1923 to 2018.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1923

    One set of grade boxes

    USDA set the first official US cotton grades in 1909, nine grades from Middling Fair to Good Ordinary, judged by eye against physical samples. The United States Cotton Standards Act became law on 4 March 1923, and that year USDA made the Universal Standards Agreement with European cotton trade organizations, so that buyers on both sides of the Atlantic graded against the same boxes. A USDA presentation hosted by the International Cotton Advisory Committee (ICAC) counts 23 foreign signatory associations from 21 countries and says the standards are used in over 50 countries. Each grade is still a box of six hand-made "biscuits" of lint, assembled only in Memphis, and a set lasts a year before its color changes.

    depts.ttu.eduicac.orgfarmprogress.com

    Credit

    USDA; European cotton trade associations

  2. 1966

    Micronaire joins the grade

    Micronaire is a reading of how easily air passes through a weighed plug of fiber, which tracks how fine and mature the fibers are. USDA added micronaire readings to the official cotton standards as a third quality factor, beside grade and staple length (the typical length of the fibers), effective 1 June 1966.

    depts.ttu.eduagresearchmag.ars.usda.gov

    Credit

    USDA

  3. 1991

    Instruments class every bale

    The high volume instrument (HVI) is a line of machines that measures micronaire, color, trash, fiber length and strength. USDA ARS says it was developed at its Clemson, South Carolina, laboratory with Motion Control, a small Dallas instrument maker, after growers in Lamesa, Texas, began raising money in the 1960s for a new testing system; Lamesa became the first US community with an HVI classing office. An integrated system could test 800 samples in an 8-hour day instead of 40, and HVI became part of USDA's official classing of all American upland cotton in 1991. In 2021 USDA's 10 classing offices classed 17.1 million samples.

    agresearchmag.ars.usda.govgovinfo.govfarmprogress.com

    Credit

    USDA ARS (Clemson); Motion Control; Lamesa growers; USDA Agricultural Marketing Service (AMS)

  4. 2004

    China tests bales at gins

    Before the reform, Chinese gins marked their own grades by eye and hand, and buyers often reopened bales to check. A plan approved by the State Council and issued on 17 December 2003 by the National Development and Reform Commission and four other agencies moved official ("public") inspection to the gin: an automatic sampler at the press takes lint from both sides of every bale, a bar-coded card goes with the sample, and fiber inspection agencies test it on HVI. It also replaced 85 kg bales with the international 227 kg bale, pressed on presses of at least 400 tonnes, which it said would cut about 60 million bales a year to about 20 million. The plan set a pilot for the 2004 season and gradual rollout from 2005. The China Fiber Quality Monitoring Center now reports inspected volumes as running totals, for example 5.45 million tonnes of Xinjiang cotton by 16 December 2025.

    ndrc.gov.cnitiger.com

    Credit

    National Development and Reform Commission; General Administration of Quality Supervision, Inspection and Quarantine (AQSIQ)

  5. 2007

    Labs checked worldwide

    Labs testing the same cotton on HVI can disagree, so the ICAC task force on Commercial Standardization of Instrument Testing of Cotton (CSITC) sends identical samples to labs several times a year and scores the results. In its fourth round trial of 2007, 58 facilities with 79 instruments on six continents took part; between labs, single strength tests varied with a standard deviation of 1.35 grams-force per tex (a unit that adjusts breaking force for fiber thickness). CSITC says round trial 2026-3 was completed in September 2026. The older ICA Bremen round trials, which Axel Drieling of the Faserinstitut Bremen (FIBRE) says date from the 1950s, had 115 labs in 2022 and reported "a relatively high scatter" between results.

    seedquest.comcsitc.orgbaumwollboerse.de

    Credit

    ICAC CSITC task force; FIBRE; Bremen Cotton Exchange

  6. 2012

    Machines take color and leaf

    USDA replaced classers' color calls with HVI color readings for the 2000 crop, after pilots in 1998 and 1999 showed the colormeter matched the official standards as accurately as classers. Leaf grade, a rating of the leaf trash in the sample, followed on 6 April 2012, after a 2009–2010 pilot in which instrument readings matched the Universal leaf standards more accurately than classers did. People still look at every sample for extraneous matter, anything other than fiber and leaf, such as bark, grass, seed-coat fragments and plastic.

    govinfo.govfederalregister.govfarmprogress.com

    Credit

    USDA AMS; National Cotton Council Quality Task Force

  7. 2018

    Plastic gets its own code

    The International Textile Manufacturers Federation (ITMF) has surveyed spinning mills about contamination since 1982. In its 2016 edition, 90 mills in 22 countries judged 23% of the cotton they used moderately or seriously contaminated, with plastic film and woven plastic among the most common contaminants, and named growths from India, Pakistan, Uzbekistan and Ivory Coast as the most contaminated. From 1 July 2018 USDA classers record plastic in a sample under two new codes, 71 and 72, instead of "other".

    LessonUSDA notes that plastic is not spread evenly through a bale, so a sample can miss plastic that is in the bale.

    itmf.orgcotton.org

    Credit

    ITMF; USDA AMS Cotton and Tobacco Program

Watch for

  1. Brazil grades by instrument

    The Brazilian Cotton Growers Association (Abrapa) says the HVI laboratories in its quality program "analyze 100% of the country's production." In July 2025, a visiting delegation of the International Textile Manufacturers Federation's spinners had reported that Brazil's "dual classification system (HVI and manual classing)" was creating market confusion and inconsistency with international standards.

    Achieved whenan outside body, such as ICAC's CSITC or a follow-up ITMF visit, reports that Brazilian bales trade on one instrument-based classification that buyers accept.

    abrapa.com.britmf.org

    Claimed by

    Abrapa, June 2026

Open

  1. 01

    Extraneous matter by camera

    An image model that reads both faces of a classing sample, or the HVI's trash-camera image, and assigns USDA's extraneous matter codes (bark, grass, seed-coat fragments, oil, spindle twist, plastic, each at level 1 or 2), scored against official classer calls.

    The brief
    Credit · unclaimed
  2. 02

    Count contamination, not opinions

    A measured contamination census: sort the foreign matter out of lint from bales of known origin at spinning mills, count and photograph each piece, and publish counts per kilogram by origin and type.

    The brief
    Credit · unclaimed

Open 01 · the brief

Extraneous matter by camera

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An image model that reads both faces of a classing sample, or the HVI's trash-camera image, and assigns USDA's extraneous matter codes (bark, grass, seed-coat fragments, oil, spindle twist, plastic, each at level 1 or 2), scored against official classer calls.

Why it is still open

USDA classing offices still have a person check every sample for extraneous matter, though instruments measure everything else. USDA ARS's Cotton Trash Identification System was compared with AMS classer calls in 2011 on 52 HVI images from one instrument; it called more extraneous matter than the classers, and the authors said larger tests were needed. In 2016 USDA ARS researchers had classers mark bark and grass on sample images; a model that was 99% right on the marked pieces was only about 77% right on pieces the machine found itself. Lab studies classify trash types when the trash is placed by hand (79.2% average accuracy over 11 types in a 2017 hyperspectral study), but I found no published test against classer calls on real classing samples at scale.

api.crossref.orgfarmprogress.comcotton.orgcotton.org

Built with what exists

A flatbed scanner or the HVI's own trash images; open detection and segmentation models such as Ultralytics YOLO; Label Studio for marking particles; samples with plastic that USDA classing offices keep for customer review.

  1. First step

    Ask a USDA classing office or a university fiber lab for 200 retained samples with known extraneous matter codes, and scan both faces of each at 1,200 dots per inch.

  2. Done looks like

    On 2,000 held-out samples with official codes, the model matches the classer's extraneous matter code on at least 90% of samples and flags at least 95% of samples coded for plastic (71 or 72), with a second classer reviewing disagreements; images, labels and code published.

Open 02 · the brief

Count contamination, not opinions

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A measured contamination census: sort the foreign matter out of lint from bales of known origin at spinning mills, count and photograph each piece, and publish counts per kilogram by origin and type.

Why it is still open

ITMF's surveys, the main worldwide comparison, record mills' judgments of each growth on a scale from "non-existent or insignificant" to "serious", and ITMF itself notes that its summary treats all 16 contaminant types as equally serious. The 16th edition, released in January 2023, is still built from 104 mills' judgments, not counts. Counts exist only for special cases: in the 2015 season two gins hand-cleaned bales already suspected of carrying module wrap and found 4 to 8 grams of plastic per bale, as Marinus van der Sluijs reported. I found no published per-origin counts per kilogram.

itmf.orgitmf.orgcotton.org

Built with what exists

1 kg lint samples from bales whose origin is on the bale tag or invoice; a light table and camera; an open detection model such as Ultralytics YOLO to count and size pieces of plastic, string, feathers and other foreign matter; a spreadsheet of results.

  1. First step

    Write the sorting method and sort 1 kg from each of 20 bales at one mill, counting and photographing every piece.

  2. Done looks like

    500 bales from at least 10 origins, each with a photo record and counts per kilogram by contaminant type, published with the method, with a second lab recounting 10% of samples to within 10%, and the census repeated a year later.