• Bharathi P S
• Kruthi Ninga Raj

Texture of knitted fabrics is an important factor for better decision making concerning their use in the production of specific types of garments. Traditional manual visual inspection for recognizing knit fabric textures faces various challenges and can be inaccurate resulting in discontent with and waste of manufactured clothing. Automating the task using a deep learning-based image identification and classification approach is a viable solution to this challenge. For accurate results, building deep learning models and starting the learning process from scratch can be computationally expensive and time-consuming. Also, a rule of thumb for deep learning based image classification is 1,000 representative images per class, which comes from the original ImageNet classification competition. In this paper, we propose transfer learning approach to recognize and classify 17 types of knit fabric texture images obtained using high resolution camera under proper lighting effects. This approach addresses both the issues related to building DL models from scratch as transfer learning uses pre-trained models that allow us to build accurate models in a timesaving way, and 1,000-image magic number goes down significantly when using such models. Three pre-trained models, Very Deep Convolutional Networks for Large-Scale Image Recognition(VGG-16), Inception-v3 and Residual net(ResNet50), are used for recognition and classification of images of 17 types of knit texture. Our models' outcomes were assessed using evaluation metrics such as accuracy, precision, recall, and F1-score. The experimental results showed that Inception-v3 achieved higher accuracy followed by ResNet50 and VGG-16.

knit fabric texture; pattern recognition; image classification; deep learning; ResNet-50; VGG-16; Inception-v3; transfer learning

"Comparison of Transferred Deep Neural Networks for Knit Fabric Texture Recognition and Classification", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.10, Issue 2, page no.f719-f729, February-2023, Available :http://www.jetir.org/papers/JETIR2302580.pdf

"Comparison of Transferred Deep Neural Networks for Knit Fabric Texture Recognition and Classification", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.10, Issue 2, page no. ppf719-f729, February-2023, Available at : http://www.jetir.org/papers/JETIR2302580.pdf