Logo Lanfrica

AlfredMichael/Himba-Vision-Public

Domaine:

healthcare

Type de record:

software
Créateur:
Alf
Hôte:
An Android-based assistive application designed to enhance contextual awareness for visually impaired individuals using advanced computer vision techniques. # **Himba Vision** 🦾👁️ *A cost-effective visual assistance solution for visually impaired individuals* ## 📌 Overview Himba Vision is an **Android-based assistive application** designed to enhance **contextual awareness** for visually impaired individuals, enabling independent navigation. It leverages **computer vision technologies** to identify objects, read text, estimate distances, and provide spatial awareness—all in real time. ## 🔍 Key Features ✅ **Offline Object Detection** using **NanoDet** ✅ **Online Panoptic Segmentation** with **Detectron2’s Panoptic FPN** ✅ **Text Recognition & Speech Interaction** 🗣️ ✅ **3x3 Grid Localization** for structured spatial awareness 📌 ✅ **Monocular Distance Estimation** for object proximity detection 📏 ✅ **Google Maps Integration** for geographical context 🗺️ ## 📱 Installation & Running the App 1️ Load the project in **Android Studio**. 2️ Connect an **actual Android device** for deployment (❌ **No emulators**). 3️ Run the application. ### ⚠️ Troubleshooting - If you encounter issues running the app, locate the **APK file** in the `release` directory and install it manually. ## 🎭 Panoptic Segmentation Setup To enable **panoptic segmentation**, follow these steps: 🔹 Create a **premium Ngrok account**. 🔹 Load the **Jupyter Notebook** found in the server-side logic. 🔹 Replace the existing **Ngrok key** with your **premium Ngrok key**. 🔹 This will expose the **Kaggle/Colab notebook hardware resources** to the app for processing. ## ⚙️ Notes 🔹 Ensure you have the necessary dependencies installed in your environment. 🔹 A **stable internet connection** is required for seamless communication between the app and cloud resources for panoptic segmentation. 🔹 If everything fails, contact me on success_fred@yahoo.com to test panoptic segmentation. ## ❌ Issues 🔹 Shadow Detection: Objects' shadows may sometimes be detected as actual objects. 🔹 Performance Variability: Latency and local processing speed depend on your device's hardwa …