OUR PROTOTYPES
Version 1
version 2
version 3
version 4
DETAILS
Hardware: Raspberry Pi 4 B, Running RPi OS Buster
OCR Software: Tesseract OCR, Post-processing
TTS Software: pyttsx3
Programming Language, Modules: pytesseract, pyttsx3, gpiozero, default modules, on Python 3.7
DETAILS
Hardware: Raspberry Pi 4 B (RPi OS Buster), mobile device (iOS only for now)
Software: Vision (iOS framework), Resnet50
TTS Software: AVSynthesizer (iOS framework)
Programming Language, Modules: flask, opencv (3rd party GitHub code), gpiozero, default modules, on Python 3.7
DETAILS
Hardware: ESP-32 CAM, Directional Haptics, SLA printing, mobile device (iOS only for now)
OCR Software: Open AI, GPT4, YOLO v11
TTS Software: AVSynthesizer (iOS framework)
Programming Language, Modules: pflask, opencv (3rd party GitHub code), gpiozero, default modules, on Python 3.7
DETAILS
Hardware: ESP-32 CAM, Directional Haptics, SLA printing, mobile device (iOS only for now)
OCR Software: Open AI, GPT4, YOLO v11
TTS Software: AVSynthesizer (iOS framework)
Programming Language, Modules: pflask, opencv (3rd party GitHub code), gpiozero, default modules, on Python 3.7
>> On device processing with raspberry pi
In Version 1, all the processing is done on the device, running Raspberry Pi OS. It uses Tesseract for OCR, as well as a post-processing algorithm. pyttsx3 is used for text-to-speech software. It was the first ever physical prototype of EyeSight and for that reason, had few supporting features.
>> Leveraging the power of mobile devices
Processing is now no longer on-device, allowing the hardware to become lighter and more affordable than ever before, while identification becomes more accurate, thanks to neural networks used. A flask server is used to stream camera data over the internet, accessed by a mobile app
>> Near Seamless mobile integration
Connecting through Wi-Fi to an app opens a lot of possibilities. It allows the physical hardware to be entirely re-imagined and made more discrete. Processing becomes extremely accurate, and software can be regularly updated. Other forms of assistance could also be developed, and this could be seamlessly linked with other software.
>> Using YOLO v11 and LLM
Optimising the hardware for user comfort, and brainstorming other possibilities for other means of assistance. Further development of the flask server and Wi-Fi stream could be done to ensure privacy.
Over the long term, development of a new OCR software could commence, specifically designed for this purpose, and a chipset made (hardware).