You are on page 1of 14

SIGN

LANGUAGE TO
TEXT
CONVERSION
CREATED BY :
JAI BADACHIYA
KARAN KATAYAL
NAVNEET JAGURI
PRIYANKA CHAUDHARY
DEVANSH TOMAR
SIGN LANGUAGE:
• Sign language is a visual
language and consists of 3
major components:
We implemented 26
symbols(A-Z) and Blank
for ASL. These symbols
are used in the
implementation of our
project.
Collection of Database:
• Dataset loading is the process of reading
and loading a dataset into memory,
typically from a file or database, for use in
machine learning or natural language
processing applications. In the context of
“sign-to-text conversion” we need an
ASL (American sign language) dataset
with high-quality images of ASL symbols
for Model testing and training.
Proposed Framework:
Convolutional Neural Networks:
● CNNs consist of multiple convolutional layers each layer containing numerous
“filters” which perform feature extraction.
● Initially these “filters” are random and by training, the feature extraction gets
better by better.
● It’s primarily used for image classification.
Methodology:
Model:
Collection of Necessary Libraries:
• Libraries are crucial tools in
software development that
provide pre-built functions
for developers to use in their
projects. They offer a variety
of benefits, including time
and effort savings, code
reliability, and optimization.
Summary for Model:
RESULTS:
• The model shows a very high
accuracy of 98.92%.. In the
context of sign language to
text conversion using a CNN
model, accuracy represents
the model's ability to
accurately classify sign
language gestures or signs
into their corresponding
textual representations.
Final Output:
Limitations:
 The model works well only in good lighting
conditions.
 Camera quality may affect the prediction.
 A plain background is needed for the model to

detect with accuracy.


Future Scope:

❖ We are planning to achieve higher accuracy even in a


noisy background by trying out various subtraction
models. And algorithms.
❖ We are also thinking of improving the preprocessing to
predict gestures in low-light conditions with higher
accuracy.

You might also like