Wednesday, 23 September 2026

Reactive Dyeing of Cotton Explained: Salt, Alkali, Fixation and Hydrolysis



Reactive dyeing is the workhorse route for producing bright, wash-resistant shades on cotton, yet it is often explained as a recipe rather than a controlled chemical process. Salt, alkali, temperature, time and wash-off do different jobs. If one is poorly controlled, a dyehouse may still obtain a dark-looking fabric but lose levelness, reproducibility or wet fastness.

The key distinction is between dye that has merely moved onto cotton and dye that has actually bonded to it. Understanding that difference makes troubleshooting far more systematic.

What makes a reactive dye “reactive”?

A reactive dye contains a coloured chromophore, water-solubilising groups and at least one reactive group capable of forming a covalent bond with cellulose. In an alkaline bath, some hydroxyl groups in cellulose become more nucleophilic and can attack the dye’s reactive group. The resulting dye–fibre bond gives properly fixed reactive shades their characteristic wash resistance.

This is more specific than ordinary attraction. Dye remaining on the surface through hydrogen bonding, van der Waals forces or physical entrapment can be removed in washing. The blog’s earlier introductions to dye–fibre interactions and dyeing methods provide useful background.

Salt ions helping reactive dye approach a cotton fibre

Salt reduces electrostatic repulsion, helping anionic reactive dye move from the bath towards cotton.

Salt mainly controls exhaustion

Most commercial reactive dyes are anionic in water. Wet cotton also develops a negative surface charge, so dye and fibre repel one another. Sodium chloride or sodium sulphate raises the ionic strength of the bath and screens that repulsion, allowing more dye to approach and adsorb on the fibre. In other words, salt primarily promotes exhaustion: transfer of dye from liquor to cotton.

Salt does not, by itself, prove that the dye has reacted. Its required amount varies with dye structure, shade depth, liquor ratio, substrate and machine. A fixed “grams per litre” rule is therefore unsafe. Dumping salt too quickly can also make the dye strike rapidly on accessible areas before circulation has distributed it evenly. Staged addition, adequate dissolution and sufficient machine turnover are practical levelness controls.

Alkali triggers fixation—and the competing reaction

Alkali raises the bath pH and activates cellulose for reaction with the dye. Sodium carbonate is widely used; other systems may use bicarbonate, caustic soda or combinations, depending on the reactive group and process. Controlled alkali dosing matters because a local pH shock can start fixation before the fabric and liquor are uniform.

The complication is that water and hydroxide ions can also react with the dye. This competing reaction is hydrolysis. Once a reactive group has hydrolysed, that dye molecule can no longer form the intended covalent bond with cellulose, although it may remain temporarily adsorbed and contribute apparent depth before washing.

Reactive dye fixation on cellulose compared with dye hydrolysis in water

Fixation and hydrolysis compete: one route bonds dye to cellulose; the other produces unfixed dye.

Exhaustion is not fixation

A nearly clear bath can be encouraging, but it does not show how much dye will remain after soaping. Exhaustion is commonly measured from the fall in dye concentration in the bath:

Exhaustion (%) = (C0 − Ct) / C0 × 100

Here, C0 is the initial concentration and Ct is the concentration at the selected time or at the end. Fixation must be assessed separately, often using colour strength before and after a defined wash-off or by analysing bath and wash liquors. A report should state its calculation basis because “fixation percentage” can mean the fixed dye as a proportion of the original dose or as a proportion of exhausted dye.

Temperature and time must suit the dye system

Raising temperature generally speeds diffusion and reaction, but it can also accelerate hydrolysis. The best profile depends on the dye’s reactive group or combination of groups. A 2019 comparison of electrolytes for one vinyl-sulphone dye, for example, found suitable exhaustion at 50 °C and lower exhaustion above that point as hydrolysis increased; that result illustrates dye-specific optimisation, not a universal set-point.

Time is equally important. The machine needs enough circulation before and during alkali addition to prevent tonal variation, followed by sufficient fixation time at the specified pH and temperature. Temperature, pH and dosing records are therefore more useful for root-cause analysis than a final shade reading alone.

Wash-off completes the process

After fixation, the material still carries unfixed and hydrolysed dye. Rinsing, neutralisation and soaping must remove it from the fibre surface and prevent redeposition. Inadequate wash-off may pass an initial visual inspection yet later cause staining, bleeding, poor wet rubbing or shade change in garment washing.

This is why colour fastness to laundering and rubbing should be assessed by specified methods, not inferred from shade depth. The result also depends on the quality of preparation: waxes, uneven absorbency, residual peroxide, hardness and metal contamination can all disturb dye access or reaction. The way cotton absorbs moisture helps explain why uniform wetting and preparation matter.

Controlled reactive dyeing with staged dosing, circulation and wash-off

A reproducible process links staged dosing and circulation to thorough rinsing and soaping.

A compact control plan for the dyehouse

Stage Main purpose Useful control
Preparation Uniform access to cellulose Absorbency, pH, residual-peroxide and hardness checks
Dye and salt Controlled exhaustion Filtered solution, staged salt and adequate circulation
Alkali Start and sustain fixation Metered dosing, verified pH, temperature and time
Wash-off Remove unfixed dye Defined rinse, neutralisation and soaping sequence

For bulk approval, a practical record includes the approved recipe and batch card, liquor ratio, addition and temperature curves, final pH, colour difference against the standard, levelness measurements and wet-fastness results. Buyers and merchandisers should also distinguish a shade correction—which can alter total chemical and water use—from a right-first-time batch.

For Indian cotton knit and woven processors operating under cost, water and effluent constraints, optimisation should aim at the lowest effective electrolyte and wash water for the specific dye range and equipment, not simply the lowest figure in isolation. Recent research has explored cationised cotton, alternative salts and non-aqueous or solvent-assisted systems. One 2024 open study reported comparable or improved exhaustion and fixation for eight reactive dyes in selected alcohol-based media, but it also emphasised dye-structure effects and the need for solvent recovery. Such approaches are promising research directions, not drop-in replacements for every production line.

The practical lesson

A reliable reactive shade is the outcome of balanced mass transfer and chemistry. Salt helps dye reach cotton; alkali enables bonding; temperature and time govern rates; hydrolysis consumes useful dye; and wash-off reveals the fixation that was truly achieved. When these functions are measured separately, corrective action becomes clearer—and quality improves together with resource efficiency.

Source acknowledgement

Technical concepts and experimental context were checked against the open-access research article “Solvent-assisted salt-free reactive dyeing of cotton fabric” (2024), the open research paper “Eco-friendly salt/alkali-free exhaustion dyeing of cotton fabric with reactive dyes” (2022), the peer-reviewed electrolyte comparison study (2019), and the review “Ecological Approaches to Textile Dyeing” (2022). Testing context follows the methods referenced in those studies, including ISO 105-C06 for laundering and ISO 105-X12 for rubbing. Readers planning production trials should use the current standard edition and their dye supplier’s technical data.

Tuesday, 15 September 2026

Recycled Polyester Explained: Bottle-to-Fibre, Textile-to-Textile and Buyer Claims



Recycled polyester is now a familiar specification in sportswear, uniforms, bags and fashion fabrics. Yet the words “made with recycled polyester” answer only one question: some feedstock was recycled. They do not tell us where that feedstock came from, how much is present, whether the finished product can be recycled again, or whether the claim has been independently traced.

The distinction matters because most recycled polyester used in textiles does not come from old garments. Textile Exchange’s Materials Market Report 2025 states that polyester represented 59% of global fibre output in 2024, that 88% of polyester was fossil-based, and that recycled polyester’s share was only 12%. Of the recycled polyester produced, 98% was still made from plastic bottles. Meanwhile, less than 1% of the global fibre market came from pre- and post-consumer recycled textiles.

Bottle-to-fibre: useful, but not a closed textile loop

In a typical mechanical bottle-to-fibre route, collected PET bottles are sorted, labels and caps are removed, and the bottles are washed and ground into flakes. The flakes are dried, melted and filtered, then converted into chips or fed directly to extrusion. Molten polymer passes through spinnerets to form filaments, which are drawn, textured or cut into staple fibre before being spun, knitted or woven.

PET bottles passing through flakes, pellets, melt spinning, yarn and knitted fabric

A simplified bottle-to-fibre sequence from collected PET packaging to polyester textile.

This route can displace some virgin polymer and gives collected bottles another use. However, it is a transfer from the packaging system into the textile system. If the resulting garment has no practical collection and recycling route, its next destination may still be disposal. Calling the product “recycled” can be accurate; calling it “circular” requires much more evidence.

A circular strategy keeps products and materials in circulation at their highest value through maintenance, reuse, refurbishment and recycling. That is why durability, repair and repeated use should not be displaced by a narrow focus on recycled-content percentage. A garment that fails early is not made circular simply because its original yarn came from bottles.

What textile-to-textile recycling changes

Textile-to-textile recycling uses textile waste as the new textile feedstock. Pre-consumer sources include cutting waste, yarn waste and rejected fabric; post-consumer sources include used garments and household textiles. Mechanical recycling opens and shreds material into fibres. For polyester, suitably clean and compatible waste may also be remelted, while chemical routes break PET into smaller molecules that can be purified and repolymerised.

Each route has limitations. Mechanical opening shortens fibres and can lower spinnability or strength, so the recovered material may need to be blended with longer fibres. Melt processing is sensitive to contamination and polymer degradation. Chemical recycling may handle some difficult feedstocks and remove certain impurities, but the result depends on the exact technology, yield, energy and chemical inputs, purification burden and local infrastructure. “Chemical recycling” is therefore a process description, not automatic proof of lower impact.

Bottle-to-garment pathway beside a closed garment-to-garment recycling loop

Bottle-to-fibre is cross-sector recycling; textile-to-textile aims to return garments to textile production.

Why an old garment is difficult feedstock

A beverage bottle is usually a relatively controlled PET article. A garment is an assembled material system. It may combine polyester with cotton, viscose, nylon or elastane; contain sewing threads of another polymer; and carry zips, buttons, interlinings, prints, pigments, coatings and chemical finishes. Dark shades, heavy finishes and unknown use histories can complicate identification and purification.

Blends are especially important. A polyester-cotton shirt cannot simply be treated as pure PET without first separating or accommodating the cellulose fraction. Even a small elastane content can disturb some recycling routes. Fibre-composition testing, dismantling, near-infrared sorting and reliable product data therefore become part of the recycling system, not peripheral activities. Readers can connect this challenge with how cotton fibre behaves and with the many surface treatments described in Textile Finishing.

Textile sorting line showing blends, elastane, trims, coatings, dyes and sewing threads

Composition, trims, colour and finishes determine whether a used garment is suitable recycling feedstock.

A buyer’s verification checklist

Question Evidence to request
What is the feedstock? Bottle or textile; pre-consumer or post-consumer; source and collection region.
How much is recycled? Percentage by mass for the product or component, not an unsupported overall impression.
How is it traced? Applicable certification scope, chain-of-custody documents and transaction evidence.
Is it physically present? Clarify physical segregation, controlled blending or an attribution/mass-balance model.
Does it still perform? Lot-wise tests for strength, dimensional stability, colourfastness, pilling and end-use needs.
Can it circulate again? Design for disassembly, compatible blends and trims, collection route and named recycler.

Certification is valuable when it verifies the claim within a defined scope, but buyers should not stretch that evidence. It does not automatically prove that the whole garment has a lower life-cycle impact, will be collected, or can be recycled in the market where it is sold. Similarly, a laboratory result should be tied to the relevant performance requirement. The earlier article on reading AATCC TM195 moisture-management results shows why a fibre claim cannot replace finished-fabric testing.

Write the claim so it can be checked

ISO 14021:2026, published in June 2026, sets principles, requirements and guidance for self-declared environmental claims and the documentation used to support them. In practical buying language, “contains 60% post-consumer recycled polyester from PET bottles by mass” is more informative than “eco-friendly recycled fabric.” If an attribution or mass-balance model is used, it should be explained rather than presented as guaranteed physical content in every individual item.

Indian textile context

India already has significant PET collection, polyester spinning and textile-processing capability. The next opportunity is not merely to increase bottle-derived yarn, but to connect garment collection, accurate sorting, dismantling, recycler specifications and traceable fibre-to-fibre output. Export suppliers can help by recording exact blend composition, reducing unnecessary material combinations, selecting detachable trims and sharing processing information that affects recyclability.

This system view is consistent with the UNEP textile circularity roadmap, which identifies shifting consumption patterns, improved practices and infrastructure investment as interdependent priorities. The commercial question is therefore not “Is recycled polyester good or bad?” It is “What material was displaced, what evidence supports the claim, how well will the product perform, and what credible next loop exists?”

Conclusion

Bottle-to-fibre recycled polyester can be a useful transition material, but it should not be confused with a closed textile loop. Genuine progress combines precise claims, traceable feedstock, durable products, design for recovery, appropriate testing and functioning textile-to-textile infrastructure. For buyers and merchandisers, the strongest sustainability claim is the one that remains clear after every term has been defined and every document has been checked.

Sources and Acknowledgement

Saturday, 5 September 2026

Knitted Garment Spirality: Causes, Measurement and Practical Control



A T-shirt can look balanced at final inspection, yet after washing one side seam may migrate toward the front while the other moves toward the back. This distortion is called garment spirality or seam twist. It is especially associated with circular-knitted single jersey used for T-shirts, vests and innerwear.

Spirality is not simply poor sewing. It reflects interactions among yarn torque, loop geometry, knitting, relaxation, wet processing, finishing and assembly. Laundering releases stresses that may be hidden in freshly finished fabric, so control must extend from spinning through final testing.

Single-jersey T-shirt before laundering and with a twisted side seam after laundering

A side seam that appears vertical before laundering can rotate as stored stresses are released.

What Spirality Actually Means

In a balanced knit, wale lines should run approximately along the garment length and courses approximately across the width. In a spiral fabric, the wales are inclined instead of being perpendicular to the courses. When garment panels are cut and joined, this inclination appears as a rotating side seam, an uneven hem or a body that no longer hangs symmetrically.

Skew describes angular distortion in a fabric. Spirality is the helical distortion common in tubular or circular knits. Seam twist is its garment-level manifestation. The terms are connected, but fabric and finished-garment tests do not necessarily answer the same question.

Why Single Jersey Is Vulnerable

Single jersey is structurally unbalanced: all face loops are oriented in the same basic manner. When a twist-lively single yarn bends into these asymmetric loops, its residual torsional energy can encourage the loop columns to lean. The effect becomes more visible when water, agitation, heat and drying allow the yarn and loops to seek a lower-energy configuration.

A systematic Hong Kong Polytechnic University study established a quantitative relationship between measured yarn twist liveliness and spirality in pure cotton single-jersey fabrics. It also showed why yarn behaviour must be measured rather than inferred only from nominal twist. Twist level matters, but fibre type, spinning route and downstream processing also affect residual torque.

Readers who want the spinning background can connect this behaviour with the role of the traveller in ring spinning. The traveller is part of the twist-and-winding mechanism, while the resulting yarn structure later influences knitting performance. The comparison of open-end and ring yarn properties is also useful because yarns with similar count and appearance may differ in structure and torque response.

Twisted cotton yarn forming leaning single-jersey knit loops

Residual yarn torque and asymmetric loop geometry can combine to incline the wale columns.

The Main Variables

Stage Risk factor Useful control
Yarn High or inconsistent twist liveliness; unbalanced singles yarn Specify torque performance, not twist alone; trial balanced or plied constructions where suitable
Knitting Loose loop length, unstable tension, feeder variation or unsuitable machine settings Hold stitch length and yarn input tension consistently; evaluate machine direction and lot variation
Wet processing Uneven relaxation, rope distortion, uncontrolled tension or inadequate finishing Allow controlled relaxation; optimise slitting, spreading, stentering and compacting
Garmenting Panels cut before relaxation, off-grain laying or inconsistent panel orientation Relax rolls before spreading; align wales and approved grain lines; keep cutting directions consistent
Testing Judging only the unwashed garment or changing the wash/dry procedure Use an agreed laundering method, number of cycles and acceptance limit for every lot

No single setting guarantees zero spirality. Compaction can improve dimensional stability, but it cannot fully neutralise a highly twist-lively yarn. A lower-torque yarn likewise cannot compensate for severe processing distortion or off-grain cutting.

Two-ply or torque-balanced yarns can reduce the tendency to rotate because opposing torque components can partly cancel. The earlier My Textile Notes explanation of EliTwist yarn gives one example of a compact, spin-twisted route. Such yarn changes must still be assessed for cost, count, hand, strength, appearance and the specific knit construction; they are options, not universal remedies.

How to Measure It Correctly

The first rule is to test the condition that matters commercially. ISO 16322-3:2021 specifies procedures for measuring spirality or torque in woven and knitted garments after domestic laundering. ISO explicitly notes that results from different procedures may not be comparable and that the method is intended for the post-laundering condition, not merely the as-manufactured garment.

AATCC now separates fabric skew and garment seam twist. AATCC TM179-2025 addresses skew change in fabrics after home laundering, while AATCC TM207-2025 is titled Seam Twist in Garments Before and After Home Laundering. Therefore, a test report should identify the exact method and edition instead of stating only “spirality tested.”

In practice, the laboratory identifies reference positions, performs the specified washing and drying sequence, conditions the specimen, lays it without stretching, and measures the required displacement or angle. ISO 6330:2021 defines domestic washing and drying procedures, detergents and ballast conditions for textile testing. Because machine type, detergent and drying route can affect results, suppliers and buyers must agree on the procedure.

A result should record direction, magnitude, number of cycles, garment size, colour, fabric lot, wash programme, drying method and conditioning. Acceptance limits are buyer- and product-specific, not universal pass/fail values.

A Practical Mill-to-Garment Control Plan

  1. Screen development yarns for twist liveliness or snarling tendency and confirm performance in a knitted trial, not only on the yarn package.
  2. Knit a representative trial using production stitch length, gauge, feeder plan, diameter and finishing route.
  3. Measure greige, dyed, finished and laundered conditions. This reveals where the distortion appears or is being temporarily masked.
  4. Relax finished rolls before spreading. Check wale alignment, bow and skew across the width and through the roll.
  5. Make pilot garments and test several sizes and colour lots after the agreed wash cycles. Garment assembly can alter how fabric distortion becomes seam twist.
  6. Release bulk only after comparing results with the approved specification and retaining a traceable record.

Textile production workflow for controlling knitted garment spirality

Reliable control links yarn selection, knitting, relaxation, finishing, cutting and post-wash verification.

For Indian clusters producing cotton T-shirts, hosiery and innerwear, checking only the finished roll or first shipment is risky. Processing conditions, large lots, subcontracted finishing and compressed schedules can change relaxation history. Pilot garments and lot-wise wash verification can prevent a visually acceptable fabric from becoming a complaint.

Conclusion

Spirality is a chain-of-process problem. Yarn torque supplies a driving force, single-jersey geometry makes rotation easier, and processing or assembly determines how strongly it appears. Finishing can manage the symptom, but robust control begins with the yarn and continues through knitting, relaxation, cutting and laundering.

The most useful question is not “Does the unwashed T-shirt look straight?” It is “After the agreed care cycle, are its wales, seams and hem still acceptable for the intended product?” That question turns spirality from a late inspection surprise into a measurable development parameter. For broader context on the final processing stage, see Textile Finishing.

Sources and Acknowledgement


Saturday, 29 August 2026

Why Accuracy Alone Is Not Enough in Textile Image Classification



A textile image classifier may report 95% accuracy and still fail on the classes that matter most. The problem is not necessarily the model; it may be the way performance is being measured.

Computer vision is increasingly used for fabric-defect detection, fibre and weave identification, garment inspection, print recognition and saree classification. Accuracy is usually the first number reported because it is easy to understand: how many test images were classified correctly? Yet textile datasets are rarely perfectly balanced, and several classes may share colours, borders, motifs, yarn effects or surface textures. Under these conditions, one overall percentage can conceal serious weaknesses.

This question extends the earlier My Textile Notes discussion on deep learning for saree texture identification. Once a model has been built, the next challenge is deciding whether its performance is genuinely useful.

Why a High Accuracy Can Be Misleading

Accuracy is the number of correct predictions divided by the total number of predictions. It is valid, but it gives every image equal weight rather than every class equal importance. Imagine a hypothetical textile test set containing 700 images from a dominant catalogue class, 200 ikat images, 80 brocade images and only 20 images from a rare handloom class. Suppose the model correctly identifies 680, 170, 45 and 2 images respectively. Overall accuracy is 897 out of 1,000, or 89.7%.

That sounds respectable. However, recall for the rare class is only 10%, and recall for the 80-image class is about 56%. The model is performing well mainly because the large classes dominate the total. A merchandiser, museum archive or provenance-research project could therefore receive a reassuring accuracy figure while the less common textiles are being missed.

Many common fabric swatches classified correctly while a small group of ornate textile swatches is misclassified.

Strong performance on a majority class can hide failure on rare but important textiles.

Read the Confusion Matrix Before Celebrating

A confusion matrix places true classes in rows and predicted classes in columns. The diagonal contains correct predictions; off-diagonal cells show the direction of errors. This is especially valuable in textiles because not all mistakes are equally informative. If two classes are repeatedly confused, the pattern may point to similar motifs, inadequate image views, inconsistent labelling, insufficient samples or a taxonomy that cannot be resolved from appearance alone.

For example, images of two saree traditions may share floral zari, contrast borders or similar catalogue photography even though their weaving structures and provenance differ. The confusion matrix reveals the pairwise problem that accuracy suppresses. It should be examined in both raw counts and row-normalised form: counts show the workload represented by each error, while row normalisation shows what proportion of each true class is being recovered.

A five-class textile confusion matrix with a strong diagonal and two off-diagonal confusion clusters.

Off-diagonal clusters identify the textile pairs that require better data, labels or features.

The Minimum Metric Set

Measure What it answers Why it matters in textiles
Per-class recall Of all true images in a class, how many were found? Exposes classes the model repeatedly misses.
Per-class precision Of all images assigned to a class, how many really belong there? Shows whether a label is being overused.
F1 score How well are precision and recall balanced? Useful when false positives and false negatives both matter.
Macro-F1 What is the unweighted average of class-level F1 scores? Prevents large classes from dominating the summary.
Balanced accuracy What is the average recall across classes? Makes minority-class weakness visible.

The official scikit-learn model-evaluation guide defines precision, recall and F-measures and explains multiclass averaging. Its documentation defines balanced accuracy as the average recall obtained on each class. In the hypothetical example above, balanced accuracy is only about 62%, far below the 89.7% overall accuracy. Weighted-F1 should also be treated cautiously because weighting by class support can again allow large classes to dominate.

Prevent Leakage Before Calculating Any Metric

Metrics are trustworthy only when the test set is genuinely independent. Textile datasets often contain several photographs of the same physical saree, garment, fabric roll or defect: full view, pallu, border, close-up and reverse side. If these related images are randomly divided, one view can enter training while another view of the same item enters testing. The model may recognise the product, background or photo session rather than learning a general textile characteristic.

A group-aware split keeps all images from one physical item together. The GroupKFold documentation describes non-overlapping groups, with each group appearing in the test set once across the folds. Depending on the research claim, an appropriate group may be a physical saree, fabric roll, production lot, motif artwork, vendor or photography session. The grouping decision should be recorded in the paper or project report.

Grouped saree images separated into training and testing sets, followed by a neural network, multiple evaluation measures and expert error review.

A trustworthy pipeline combines group-aware splitting, several metrics and expert review of difficult image pairs.

Match Evaluation to the Real Decision

The best metric depends on how the system will be used. For automatic catalogue tagging, precision may be important because a wrong public label damages trust. For archive retrieval, recall may matter more because missing a relevant textile is costly. For an expert-assistance tool, top-k accuracy can be useful: the correct class may be accepted if it appears among the model’s three most likely suggestions. Top-k performance, however, should complement rather than replace top-1 and class-wise results.

Error cost also varies. Confusing two colourways within one construction is not equivalent to assigning an unsupported provenance or craft identity. Visual appearance alone may not prove fibre composition, weaving technique, authenticity or geographical origin. A textile expert should review recurring error pairs and decide whether better images, microscopic structure, metadata or a revised label hierarchy is required.

This evaluation logic also applies beyond image classification. Readers interested in data-driven textile decisions may connect it with Python modelling for cotton yarn quality, optimising yarn quality through raw-material parameters, and marker efficiency as an optimisation problem. In every case, a model is useful only when its evaluation reflects the actual industrial or research decision.

A Practical Reporting Checklist

  1. Report the number of images and physical groups in each class.
  2. Explain how train, validation and test sets were separated.
  3. Provide accuracy, macro-F1, balanced accuracy and per-class precision and recall.
  4. Show raw and row-normalised confusion matrices.
  5. Review the most frequent error pairs with a textile expert.
  6. Test on images from new products, lots, vendors or capture conditions.
  7. State what the model cannot establish from an image alone.

Conclusion

Accuracy is a useful starting point, not a complete verdict. Textile image datasets combine class imbalance, fine visual differences, repeated views and domain-specific error costs. A reliable evaluation therefore needs class-wise metrics, a confusion matrix, group-aware splitting and expert interpretation. The most important question is not “How high is the accuracy?” but “Which textiles does the model recognise, which does it confuse, and would those errors be acceptable in the intended application?”

Sources and Acknowledgement

Total Pageviews