NETRA

Netra is wearable device that helps visually impaired people navigate by providing speech descriptions of the world around them. This repository implements Netra Daemon that runs on a Raspberry Pi device connected to the standard PiCamera module. To provide real-time environment descriptions Netra-Daemon needs a Netra Vision server running on a computer. Netra Vision implements a Convolutional-Recurrent Neural network that converts images to text descriptions. Rest of the modules requre Google Vision API authentication to work.

Netra Prototype Implementation

Setup

To use describe mode

1. Start the Netra Vision Server by following the instructions specified in /~https://github.com/subho406/Netra-Vision
2. Change the global IP variable in netra_daemon.py to the IP address of the Netra Vision server.

To use text, raw and face modes

1. Follow the instructions in https://cloud.google.com/vision/docs/auth to set up Google Vision API authentication.

Running

python3 netra_daemon.py MODE

Available Modes

1. describe: Environment Description
2. text: Text Recognition
3. raw: Label and Context Identification
4. face: Face Detection

Requirements

1. Google Vision API (Required for text,face and raw modes)
2. Netra Vision Server (/~https://github.com/subho406/Netra-Vision)