2025, Vol. 10, No. 1. - go to content...

Permanent address of this page - https://kostumologiya.ru/en/13ivkl125.html

Метаданные этой статьи так же доступны на русском языке

Full article in PDF format (file size: 1.2 MB)

For citation:

Karshakova L.B., Dobrovolskaya N.A. Recognizing items of clothing using a neural network. Journal of Clothing Science. 2025; 10(1). Available at: https://kostumologiya.ru/PDF/13IVKL125.pdf (in Russian).

Recognizing items of clothing using a neural network

Karshakova Lidiia Borisovna
Russian State University named A.N. Kosygin (Technologies. Design. Art), Moscow, Russia
E-mail: karshakova-lb@rguk.ru
ORCID: https://orcid.org/0000-0003-2158-2508
RSCI: https://elibrary.ru/author_profile.asp?id=745723

Dobrovolskaya Nadejda Anatolevna
Russian State University named A.N. Kosygin (Technologies. Design. Art), Moscow, Russia
E-mail: nadejda_dobro@mail.ru
ORCID: https://orcid.org/0009-0002-7481-0893
RSCI: https://elibrary.ru/author_profile.asp?id=1265475

Abstract. This article presents an approach to the development and implementation of a neural network in Python aimed at automatic clothing recognition based on images from a dataset consisting of six categories. Clothing recognition can be useful for creating a clothing database, where the description of clothing will be the key to finding a specific type of clothing in this database. The main purpose of the research was to create a machine learning model capable of classifying textiles.

As part of the work, a dataset was used that includes images of clothing divided into categories of outerwear: jackets, shirts, polo shirts, T-shirts, T-shirts and hoodies. Despite careful data preprocessing, which involves resizing images and normalizing color values, the recognition accuracy of clothing categories was only 16 %. This was due to the insufficient dimension of the data set and the similarity of the two categories, which made it much more difficult to distinguish them.

The neural network code includes several key components: a pre-trained VGG16 network, two fully connected layers (Dense), Flatten and Dropout layers. The network architecture was configured in accordance with the best practices, but the limitations of the source data had a negative impact on the results.

The experimental results emphasize the need to use more extensive and diverse data sets to improve classification accuracy. The article discusses possible ways to improve the model, including data enrichment methods that can help improve the accuracy of the model.

The conclusion of the article focuses on the challenges in the field of deep learning for object recognition and classification and emphasizes the importance of adequate data selection and preparation for the successful operation of neural networks in real-world tasks.

Keywords: clothing recognition; artificial intelligence; neural network; neural network development; pre-trained neural network; neural network learning problems; Keras and TensorFlow; convolutional neural networks; clothing databases

Download article in PDF format