DIGITAL METHOD TO TALK WITH MACHINE USING IMAGE PROCESSING AND MACHINE LEARNING TECHNIQUES

Talking to a Machine is a Machine Learning Algo- rithmthat can answer inquiries regarding an image's content. A statement, a phrase, or a single word might be used as the response. Our model has four parts: an LSTM for extracting the question representation, a CNN for extracting the visual representation, an LSTM for storing the language context in a response, and a fusing component for combining the input from the first three componentsand generating the answer [2] [3]. To train and assess our mQA model, we create a dataset called Freestyle Multilingual Image Question Answering (FM-IQA). It has approximately 150,000 photos as well as 310,000 freestyle question-answer pairs with English translations. Human judges use a Turing Test to assess the quality of our mQA model's produced responses on this dataset. [1] In particular, we combine human responses with our model. Thehuman judges must be able to tell the difference between our model and the human. We provide techniques for keeping an eye on the quality of the evaluation process. Human judges cannot identify ourmodel from humans in 64.7 percent of situations, according to the experiments. The overall average is 1.454. (1.918 for human). The FM-IQA dataset, as well as the specifics of this project.