TRANFORMERS BASED MULTI-LABEL IMAGE CLASSIFICATION AND NAMING - A LITERATURE SURVEY

Computer Engineering

Machine learning (ML), a subset of artificial intelligence (AI), has seen a sharp rise in utilisation in recent years. Through the use of machine learning (ML), software programmes can predict results more accurately without having to be explicitly trained to do so. Speech recognition, face detection, and other applications use classification, a type of ML that divides input data into various types. Image classification is taken into account here. Multi-label classification is a technique for determining a set of labels based on the characteristics and items visible in the provided image. Transformers are categorization models that have been applied to NLP (NLP). It is now utilised in picture classification. After feature extraction, the data is passed to the transformer. It will then be able to anticipate the labels in the image. In this essay, we'll talk about Vision Transformer (ViT). It is a paradigm for multi-label image classification that can take advantage of the intricate relationships between labels and visual attributes. The Multilayer Perceptron (MLP) idea is employed in this instance to extract features from the photos. This method trains the vision-transformer to anticipate the labels that will be hidden from view based on the input that is provided. The CIFAR-10 dataset was used in this case for training. The label mask training objective is the method's essential component. During training, this model can explicitly describe label state. As a result, this model can be used to recognise wild animals, medical images, and other things. In terms of prediction accuracy, this model can operate at an efficiency of roughly 84 percent.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.