A multifunctional assistive system engineered to support visually impaired individuals through advanced computer vision and navigation technologies is presented in this paper. The functionalities of the framework are divided into outdoor and indoor modules, each addressing specific challenges in accessibility. For outdoor navigation, the system integrates Google Maps API for geolocation and route planning, coupled with a monocular depth estimation pipeline based on convolutional neural networks to enable real-time obstacle detection and avoidance. Indoor exploration capabilities include a suite of deep learning-based recognition models tailored for object classification and with potential applications towards the semantic understanding of the environment. These include CNN-based modules for color recognition, clothes recognition, food classification via transfer learning on curated datasets, thermal imaging for hot object identification, and object detection models (e.g., YOLOv5) for real-time grocery identification in retail environments. The system is designed for deployment on embedded hardware platforms with optimized inference performance, offering a lightweight yet robust solution for situational awareness. The tests performed show the system’s efficacy in diverse real-world scenarios, highlighting its potential to significantly enhance the mobility, safety, and independence of visually impaired users.