Visual question answering model for plant disease identification

Abstract

The notable achievements in AI tasks owe their success to the natural language processing (NLP) domain with Large Language Models (LLM) and led to the emergence of new research directions in Deep Learning. The Visual Question Answering (VQA) task has garnered considerable attention owing to its promising results obtained through the use of pre trained LLMs. Here we are investigated a VQA as a domain specific expert system for domain specific knowledge representation and extraction. We have implemented a novel approach for plant disease identification, an expert-level task, utilizing fine-tuning LLM. The VQA technique has been utilized as a means of knowledge extraction, making it more accessible to non-expert users. We proposed a new VQA architecture that employs a fine-tuned GPT2 model for domain- specific knowledge representation, with the aim of enhancing both explicit and implicit reasoning in the context of plant disease question answering.