Published Paper Details:

This project represents an initiative to enhance accessibility and inclusivity for individuals grappling with medical conditions like visual agnosia, a neurological condition characterized by difficulties in recognizing and interpreting visual information. The technical foundation is built upon a sophisticated two-stage architecture. Firstly, the Image Encoder leverages the CLIP encoder, to extract high-level features from images. These features serve as a rich representation of the visual content and are subsequently passed to RNN. The RNN employs LSTM network well-suited for sequential data processing. The LSTM is responsible for decoding the extracted image features into coherent and descriptive textual captions. Furthermore, an integral part of this project is the integration of the gTTS (Google Text-toSpeech) library, which introduces text-to-speech capabilities. This in addition lets the transformation of retrieved textual captions into spoken words, thereby creating a comprehensive and accessible experience for individuals with visual agnosia. The deployment of gTTS not only facilitates generating of audio descriptions but also enables users to customize speech speed, language preferences, and output formats. The system's overarching objective is to provide individuals with visual agnosia a robust and adaptable toolset for interpreting visual content. By combining image feature extraction with sequence generation and auditory synthesis, this project aims to bridge the gap in understanding visual stimuli, empowering users with detailed textual descriptions and spoken narratives. The intricate interplay of deeplearning method, neural network architectures, and library integrations underscores the project's technical complexity and potential impact on lives of individuals facing challenges in visual recognition.