The proposed work introduces an innovative approach to classifying glitches in gravitational wave detector using the Swin Transformer model, a cutting-edge deep learning architecture. Gravitational waves, predicted by Albert Einstein in 1916, are ‘ripples’ in space–time caused by violent cosmic events such as colliding black holes and supernovae. Interferometers like LIGO and Virgo play a crucial role in detecting these waves by measuring the stretching and squeezing of space–time. Data from ground-based gravitational-wave detectors contain numerous short-duration instrumental artifacts, called ‘glitches’. Glitches can obscure or mimic a true gravitational-wave signal. Utilizing the Gravity Spy Zooniverse project dataset was derived from LIGO observations. The work focuses on categorizing glitches into 22 distinct classes, such as scratchy, blip, koi pond, and more. The dataset underwent meticulous preprocessing, including axis removal and partitioning into training, validation, and test sets. The work demonstrates the Swin Transformer's remarkable performance, achieving an accuracy of 0.9016, a loss of 0.5605, with an F1 score of 0.8973, precision of 0.9015, and recall of 0.8989. Additionally, the model's inference time of 140.038763 s ± 2.834662 s and an inference rate of 0.029569 s/sample ± 0.000599 s/sample demonstrate its practical efficiency in real-world applications.