DevMesh

Image Recognition with eBPF using Computer Vision

1 0
  • 0 Collaborators

Integrate Privacy and Discoverability in AI Systems by Providing a Private and Secure Data Transfer through Computing and Networking Systems ...learn more

Project status: Published/In Market

oneAPI, Internet of Things, Artificial Intelligence

Groups
DeepLearning, Artificial Intelligence Europe, Movidius™ Neural Compute Group, Intel AI DevCamps, oneAPI Showcase, Intel Deep Learning Group

Intel Technologies
oneAPI, Intel Python, OpenVINO

Docs/PDFs [1]Code Samples [1]Links [1]

Overview / Usage

Problem

The problem of processing personal data through Computer Vision and AI systems have caused issues with Data Protection

Impact Metrics

  • Improving Monitoring through Packet Sniffing by introduction of a Networking Protocol
  • A Hand-Shake Protocol for initiating the Computer Vision Results through a Vision Exchange Request
  • Less Power Usage due to the Usage of VEXF Protocol

Frameworks/Tools/Technologies

  • C++, oneAPI, Python, Python-HPC and oneAPI supported Python

Assumptions, constraints, and solution decision points

  • Computer Networking Solution with C and Python, as there are popular packages such as pcap ●eBPF Observability Solution demonstrated as integrated with C and Python code

Ease of Implementation

  • A Python Package implementing VEXF Frames will only involve Code Quality, Performance-based Computing, Conducting tests on Ethernet and WiFi interfaces
  • A HTTP Request-Response Cycle is implemented through a Database, Web-Server and Edge-Computing platforms such as an MQTT Request Broker
  • Testing would involve distributing the solutions to various providers who will implement this Package, for Discoverability or Protecting their Personal Data and Intellectual Property Related Information
  • Open Source Helper Computer Programs could assist the Package to reach the Market

Additional Scalability / Usability

  • Computing Systems such as Cloud and Edge, Sensors such as WiFi Sensors, Packet Sniffers for capturing Frames in Wireshark, Software such as HTTP Monitoring Tools like Telerik Fiddler
  • Typically, any interference can be represented using Data Structures involved in Reading and Transmitting a Vision Exchange Format Packet

Methodology / Approach

Private & Secure Data Exchange

Private & Secure Data Exchange is achieved through a VEXF (Vision Exchange Format) Protocol, that sends information through Packet Metadata, suitably in an Open-Ended Problem of your Use Case.

  1. Discoverability: Through HTTP
  2. Private & Secure: Through Networking Frames

In an Open Area, the Construction Sites must support these to enable more Safety at Site

  1. Turn off / on certain Regions at Site, for your monitoring using Vision as a Sensor
  2. Person Safety and Monitoring will be Enhanced depending on an Operational Machinery

​

Technologies Used

Python

PyTorch

Intel NNCF (Neural Network Compression Framework) Library for Initial Inference

PaddleOCR (Text Recognition)

EasyOCR (Text Recognition)

TRDG (Synthetic Dataset Generator)

Intel OpenVINO for Text Recognition using Text Spotting Model

Plotting Libraries

Jupyter Notebook

MMOCR (Text Recognition)

bcc C++ Library for eBPF

RoboFlow Python API

Numpy Numerical Library

Scikit-Learn and MNIST Dataset

FCNTL, Termios and IOCTL Libraries

Documents and Presentations

Intel_oneAPI_Hackathon_Grand_Finale_2.pdf

Repository

https://github.com/KlinterAI/image-recognition-with-privacy

Comments (0)