A new study led by Warren Jasper, a professor at Wilson University in the United States, demonstrated how machine learning can help reduce waste in textile manufacturing by improving the accuracy of color prediction during the dyeing process.
The study, entitled “Controlled Research on Machine Learning Applications for Predicting the Color of Dry Fabrics from Wet Samples: The Effects of Dye Concentration and Pressure” addresses one of the industry's long-standing challenges.
Textiles are usually dyed while wet, but often change color as they dry. This makes it difficult for manufacturers to determine the ultimate appearance of the material in production. According to a paper co-authored by Samuel Jasper, color changes are further complicated by the fact that it is impossible to generalize data from one color to another, as the color changes from wet to dry are nonlinear and differ in different shades.
“The fabric is dyed while wet, but the target shade is dry and wearable. That means if there is an error in the color, you don't know until the fabric has dried. More fabrics are dyed for a long time while you wait for that drying to occur. Warren Jasper.
To address this, Jasper has developed five machine learning models, including neural networks specially designed to handle the nonlinear relationship between wet and dry conditions. The model was trained with visual data from 763 fabric samples stained in different colors. Jasper pointed out that each staining process takes several hours and data collection becomes a time-intensive task.
All five machine learning models outperform traditional non-ML approaches in predicting final fabric colours, but neural networks have proven to be the most accurate. A low Ciede2000 error of 0.01 and a median error of 0.7 were achieved. In comparison, other machine learning models showed error ranges from 1.1 to 1.6, while the baseline model recorded errors of 13.8.
The Ciede2000 formula is a standard metric for measuring color differences, and in the textile industry values above 0.8-1.0 are generally considered unacceptable.
By allowing for more accurate prediction of the final fabric color, neural networks help manufacturers avoid expensive dyeing mistakes and reduce material waste. Jasper has expressed his desire to see similar machine learning tools be adopted more widely across the textile sector to support efficiency and sustainability.
“We're a little behind the curve of textiles. The industry has begun to move further towards machine learning models, but it's very slow. These types of models can provide powerful tools to reduce waste and increase the productivity of continuous dyeing, which accounts for more than 60% of dyeing.
A study led by Warren Jasper shows that machine learning can reduce fiber dyeing waste by accurately predicting the color of dry cloth from wet samples. The neural network model trained on 763 samples achieved near perfect accuracy and avoided expensive errors. Jasper encourages wider adoption to increase the sustainability and efficiency of continuous dyeing.
Fibre2Fashion News Desk (Hu)
