Automatic Detection and Recognition of Fish to Help Visually Impaired People

Abstract

This paper explores fish identification for blind people utilizing multi-image recognition including deep learning techniques. Our system was built using Speech Recognition to support blind people in knowing about various fish in the market or somewhere else. According to our preliminary study, the majority of blind people in Bangladesh face difficulties purchasing various products, especially fish, from the market. However, there is no automated device available that can identify the fish and offer an interpretation. For this generosity of spirit, we created a system for visual imperial people that will automatically identify and explain the fish using speech recognition. We employed Deep Learning (DL) resources like Matploatlib, TensorFlow, Keras, and others throughout the process. We utilized TensorFlow for image preprocessing and classification. We used the most popular image processing algorithm, Convolutional Neural Network (CNN), to evaluate the reliability of our work. We tested three CNN models to see which one offered the best performance. We compared all of the models before identifying multiple fishes with the model that provided the highest precision and efficiency.