An Automatic System for Identifying and Categorizing Tribal Clothing Based on Convolutional Neural Networks

The quantity of internet businesses providing tribal clothes is constantly increasing, and people tend to exaggerate how often they shop at such sites. However, we are concerned about the authenticity of the outfits. The study recommends using Convolutional Neural Networks (CNN) to automatically identify and categorize authentic images of particular tribal dresses used by some Bangladeshi tribes into predetermined categories. The study's impetus comes from the expansion of commerce and the desire to spread these traditional clothes over the globe. In order to categorize the clothing, we obtained images from actual tribal residences, shops, and a few online marketplaces. To that end, we made an effort to provide a dataset we've labeled “TribalBd,” which has 680 samples, including six different classes. Then, use the YOLOv5, YOLOv6, and YOLOv7 models to put these datasets for detection and classification on our CNN. As a means of evaluating the efficacy of our model, we have experimented with a number of different CNN topologies and tweaks. We put the model through its tests with YOLOv6 and YOLOv7. YOLOv5 achieved the best results among these models. The final result shows that the YOLOv6 model gives 86.24%, the YOLOv7 model gives 71.28% accuracy whereas YOLOv5 gives 89.97% accuracy in classifying the images in the training and testing sets which are best compared to the other two models.