Hindi Image Captioning Using Deep Learning
Soumyarashmi Panigrahi, Jaydev Sutar, Dibya Ranjan Das Adhikary · 2024
In the era of Artificial Intelligence (AI), the real-world need for Image Captioning (IC) lies in making visual content accessible, understandable, and useful across diverse applications, from aiding the visually impaired to enhancing e-commerce and enriching social media engagement. An image needs to have a description, which is the idea behind image captioning. There has been a lot of study already done in this area, but it has primarily concentrated on creating image descriptions in English language because the majority of the datasets for existing image captions are in English language, and language must not create a constraint on the IC generator model. So we have tried to generate the caption in Hindi language. Hence, we used Flickr8k with Hindi caption as our dataset. We have incorporated a diverse strategy to achieve the goal by combining Recurrent Neural Networks (RNN) and Convolutional Neural Networks (CNN). To achieve and maintain the knowledge for a longer period of time, we have also utilized Long Short-Term Memory (LSTM), a sort of RNN. Additionally, many applications of the same have also been covered. The results obtained have significant role to the understanding and management of IC in the practical field.