DEVELOPING CNN-AUGMENTED MODELS TO PREDICT CIELAB OUTCOMES POSTBLEACHING OF DENIM GARMENTS

Denim garment production demands efficient design processes to minimize waste, costs, and production delays. Bleaching, among other finishing processes, holds paramount importance due to its numerous variables and substantial impact on product value. Artificial neural networks have great potential to achieve superior performance in anticipating various process outcomes. Their parameterized structure effectively captures non-linear relationships between input features. This study aims to effectively predict fabric outcomes by developing an artificial neural network (ANN) model supported by convolutional neural networks (CNN) to provide additional features derived from raw and semi-processed fabric images. The study represents a comparison of CNN powered models with a common predictive ANN as base model. Competing models incorporate various process variables and fabric properties, such as dying number and elasticity to predict changes in denim CIELab properties after bleaching. The process features of the model are the number of bleaching cycles, total process time, and concentration of sodium hypochlorite (representing the total amount of chemical used). The mean absolute percentage error is used as the performance measure between predictions and desired outputs. This research plays a significant role in enhancing agility in denim production by providing businesses with more efficient approaches to digitized denim bleaching and Research and Development processes in the textile industry.