Scribd
Upload
6 views9 pages

Batch66 First Review

Uploaded by

Hari Krishna
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
6 views9 pages

Batch66 First Review

Uploaded by

Hari Krishna
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 9

RAGHU ENGINEERING COLLEGE

AUTONOMOUS
DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

Raspberry Pi Based Text and Face Recognition for Visually


Impaired

PROJECT GUIDE: PRESENTED BY:


I. Krishna Rao, • Vishal Paul (21985A0432)
Associate Professor, • S. Abhinai Yadav (21985A0444)
• V. Sai Abhishek (20981A04M1)
Dept. ECE • V. Shahil Rao (21985A0431)
OBJECTIVES:
⚬To develop an advanced face recognition algorithm utilizing OpenCV and
the HAAR features.
⚬To integrate an offline Text-to-Speech system, eliminating the need for
internet connectivity and providing seamless audio feedback for enhanced
autonomy.
⚬To develop a text recognition(OCR) mechanism using Pytesseract for
Visually Impaired.

⚬To implement Bluetooth connectivity for audio output, replacing wired


headsets to enhance user comfort and usability.

Simulation tests have been conducted in a PC to validate the effectiveness and


reliability of the system in providing essential assistance to visually impaired
users.
SIMULATION METHODOLOGY
The software tools used for simulations are:
• Python 3.11, libraries like pyttsx3, OpenCV and face_recognition, Pytesseract.
• Visual Studio Code (VS Code) for writing and debugging the project codebase,
providing a versatile and user-friendly development environment.
• A personal computer served as the hardware platform for conducting simulations
and testing the implemented algorithms. It provided the necessary computational
resources for processing image data and executing the simulation environment.

Dataset and Face Recognition: An algorithm for face recognition was made based on
dataset comprising images of popular celebrity face (Virat Kohli) is utilized for
training and testing the face detection component. These images serve as the basis for
extracting Haar features, which are subsequently employed to detect faces in real-
time using a webcam feed. Subsequently another algorithm was written for Optical
Character Recognition (OCR).
SIMULATION RESULTS (FACE RECOGNITION)
Face recognition algorithm successfully recognising the face of Virat Kohli,
this image wasn’t used in the training dataset.

Datatset Used

Output
SIMULATION RESULTS (TEXT RECOGNITION)
Text recognition algorithm successfully recognized the text and gave audio
output to the user.

Webcam Capture Output


COMPARISIONS WITH EXPECTATIONS
⚬The output of the face detection algotithm was even better than expected
as it was able to detect faces even with specs, with minimal training.
⚬As seen in our example, we only used 4 images of the person, to make the
program light and to avoid unnecessary processing in the computer.
⚬Still the program was able to detect the person almost everytime with
accuracy.
⚬Coming to the working of Optical character Recognition Program(OCR), it
was able to detect the written text from the webcam. It has high accuracy
only when the text is large in size or when a very high quality camera is
used, which in turn will increase the cost of the product. Hence, we
maintain a trade-off between cost and its ability to produce highly accurate
results. The results are satisfactory to be used in a raspberry pi.
FUTURE STEPS
⚬Transitioning the developed face recognition algorithm, validated
through laptop simulations, to the Raspberry Pi platform. This
involves optimizing the codebase to ensure seamless execution and
efficient resource utilization on the embedded system.
⚬Implementing an offline TTS solution directly on the Raspberry Pi
to facilitate instantaneous conversion of detected text into auditory
feedback. This step enhances user experience by eliminating the
need for internet connectivity and ensuring uninterrupted audio
assistance.
⚬Leveraging wireless Bluetooth technology to establish seamless
communication between the Raspberry Pi-based system and the
user. This enables the relay of pertinent information, including
detected faces and surrounding text, to the visually impaired user
via a wireless earpiece, enhancing mobility and user comfort.
REFERENCES
[1] A. Suresh, C. Arora, D. Laha, D. Gaba, and S. Bhambri, "Intelligent smart glass for visually
impaired using deep learning machine vision techniques and robot operating system (ROS)," Adv.
Intell. Syst. Comput., vol. 751, pp. 99–112, 2019, doi: 10.1007/978-3-319-78452-6_10.
[2] "About - OpenCV." https://opencv.org/about/ (accessed Jan. 11, 2024)
[3] Raghad Raied Mahmood, Dr. Majid Dherar Younus and Dr. Emad Atiya Khala, "Real Time Object
Detection for Visually Impaired Person", Annals of R.S.C.B., ISSN:1583-6258, Vol. 25, Issue 4, 2021,
Pages. 14725 - 14732
[5] https://www.tensorflow.org/lite/examples/object_detection/overview (accessed Feb. 14, 2024),
"Python-tesseract, pytesseract 0.3.10"
[5] https://cocodataset.org/#homehttps://docs.opencv.org/3.4/db/d28/tutorial_cascade_classifier.html
(accessed Feb. 13, 2024), “OpenCV: Cascade Classifier”
[6] https://www.python.org/doc/l (accessed Feb. 10, 2024), “Official Python Documentation”

You might also like