This project generates a caption for an image, translates it into Nepali, and reads it aloud. It supports image upload and camera input. It uses a BLIP model for captioning, Google Translate for translation, and gTTS for speech output.
This system recognizes handwritten Nepali characters using a CNN model (TinyVGG). It classifies 58 classes including vowels, consonants, and numerals. The model achieves around 83.3% accuracy on the dataset.