Image classification method and apparatus thereof based on vision-language model using decision graph

An image classification apparatus based on a vision-language model, the apparatus comprising:

• a processor with a computing function;
• a memory connected to the processor;
• a data input circuit receiving a data list including at least one image;
• a prompt input circuit specifying classification of an image and receiving a prompt including words classified into at least one class;
• a decision graph searching circuit reading a graph defining a classification order of at least one class, selecting a node of the graph, reading class information corresponding to the node, and returning a classification result if the node is a leaf node;
• an image classification circuit classifying an image based on the prompt and class information;
• and
• a natural language processing circuit extracting a word included in the prompt and classified into at least one class,
• wherein the image classification circuit is configured to:
• receive, from the natural language processing circuit, a word corresponding to the class information;
• read, as a list of classification target images, each of:
• the data list during an initial execution of the classifying by the image classification circuit, and
• a result of a previous classification of the image during repeated executions of the classifying by the image classification circuit;
• select at least one image corresponding to the selected word among images included in the list of classification target images;
• and
• output the list of selected images to be stored in the memory.