Published Paper Details:

This work presents the development of an AI-based image translation web application designed to recognize objects within an image and provide translations in Hindi, English, and Kannada. The system integrates advanced computer vision and natural language processing techniques to deliver accurate and real-time translations. Object detection is carried out by YOLOv5s (You Only Look Once version 5 Small) which is an extremely effective deep learning model developed for real-time object identification. OpenCV applies its power for operations such as image preprocessing- image resizing, normalization, and noise reduction, ensuring high-quality inputs for the model. The system utilizes TensorFlow for processing in neural networks and PyTorch for training and fine-tuning the deep learning models involved in object recognition. To enhance its functionality, the application also employs Tesseract OCR (Optical Character Recognition) to get the text from images that shall be translated into the selected languages. With a user-friendly interface, users can upload images and receive immediate translations of recognized objects and text in Hindi, English, or Kannada. This AI-powered image translation tool offers a practical solution for overcoming language barriers, benefiting diverse user groups in fields such as education, travel, and everyday communication.