Assistive Object Recognition and Tracking System for the Visually Impaired Using CNN
1 Prof. Shwetha L, Assistant professor, 2 Gowthami T S, Student, 3 Madhushree B, Student,
4Reshmashree M C, Student
Department of Artificial Intelligence and Machine Learning Engineering
G Madegowda Institute of Technology, Bharathinagara, Maddur TQ,
Mandya dist-571422, India
Abstract - The visually impaired community faces significant challenges in perceiving surrounding objects, reading textual information, and safely navigating through unfamiliar environments. Although assistive tools such as white canes and guide dogs provide partial support, they lack the capability to deliver real-time environmental awareness and contextual understanding. Recent advancements in computer vision and deep learning have opened new possibilities for developing intelligent assistive systems that can enhance independent mobility and situational awareness for visually impaired individuals.
This paper proposes an Assistive Object Recognition and Tracking System for the Visually Impaired using Convolutional Neural Networks (CNN). The proposed system utilizes deep learning–based object detection, optical character recognition (OCR), and road or lane detection models to analyze the surrounding environment through a mobile device camera. The system detects and identifies objects, reads textual information such as signboards and labels, and recognizes road or lane patterns to support safe navigation. Spatial analysis techniques are applied to estimate object direction and distance, enabling meaningful feedback.
The system operates in both offline and online modes. Offline processing is achieved using optimized TensorFlow Lite models embedded within the mobile application, ensuring reliable operation without internet connectivity. When online connectivity is available, the system retrieves additional contextual information related to detected objects to enhance user understanding. Voice-based interaction and text-to-speech output are integrated to provide real-time auditory guidance, making the system accessible and user-friendly.
By combining CNN-based vision models, spatial awareness, and voice interaction, the proposed assistive system improves environmental perception, enhances navigation safety, and promotes independent living for visually impaired users.
Key Words: Assistive Technology, Object Recognition, Convolutional Neural Networks, Computer Vision, Visual Impairment, Deep Learning, Voice Assistance
