Science Fair Project - Teachable AI - Final Normal Text

1) The experiment aims to investigate how well an AI algorithm can recognize and categorize facial expressions by comparing its performance to humans. 2) The hypothesis is that AI can recognize human emotions on faces with high accuracy given large training data and advanced machine learning algorithms. 3) The experiment demonstrates that the AI achieved an overall accuracy of 90% in recognizing happy and sad faces, though performance was influenced by lighting and expressions.

Uploaded by

bahaa

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

69 views6 pages

Science Fair Project - Teachable AI - Final Normal Text

Uploaded by

bahaa

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 6

3.

Despite advancements in artificial intelligence,

accurately recognizing and interpreting human facial
expressions remains a significant challenge. This
project aims to investigate the effectiveness of an AI
algorithm in recognizing and categorizing facial
expressions and to determine whether AI can achieve
comparable or superior performance to humans in this
task.

We hypothesize that AI can recognize human emotions

on faces with a high degree of accuracy, given the
availability of large amounts of training data and
advanced machine learning algorithms.

1) Pencils and coloring pencils

2) Scissors
3) Paper
4) A computer with a webcam and access to the
internet
5.

Artificial Intelligence (AI) is a rapidly growing field of

computer science that aims to build intelligent tools
capable of performing human-like tasks. Machine
Learning (ML) is a sub-division of AI that develops tools
that can learn from data and improve their performance
over time without human intervention. ML algorithms use
large amounts of data to discover patterns and make
decisions, which makes them highly effective at handling
complex, changing environments and finding patterns that
may be difficult for humans to detect. However, ML does
have limitations that need to be considered. Firstly, it
requires large and neutral data sets to learn from, and it can
be computationally expensive. Additionally, it is
susceptible to systematic errors and requires validation and
verification to ensure accuracy and reliability. Despite
these limitations, ML is a powerful and rapidly evolving
tool with many potential applications across various
industries. As technology advances, it is important to
remain aware of its capabilities and limitations and to
approach its use with appropriate caution and scrutiny.
1) Creating Learning Data: We drew 40 different
happy and sad faces to be used as learning data.

2) Creating Testing Data: We asked friends and

family to draw happy and sad faces to be used as
testing data.

3) Preparing the AI: We prepared the AI for the

experiment and labeled the classes.

4) Uploading the Learning Data: We uploaded the

learning data which consists of 20 pictures into
each class.

5) Training the AI: We had the machine learning

model train and generate predictions on new data.
The training process involved feeding the data to
the model, which would then use various
algorithms to learn from the patterns and
relationships within the data.

6) Testing the AI: We provided the AI with a dataset

of 20 images that the AI was not trained on and
measured how well it could recognize the faces in
those images.
7) Analyzing the Data: We collected and analyzed
the results obtained from the AI’s output and
identified any patterns or trends in its performance.
We also created a confusion matrix to organize the
results and communicate the AI's performance.

Our experiment demonstrated that AI is capable of

accurately recognizing faces with a high degree of
accuracy, achieving an overall accuracy of 90%.
Analysis of the data revealed that the AI's performance
was influenced by factors such as lighting conditions
and facial expressions, but adjustments to the training
data and algorithms were able to improve performance.
The potential for AI to accurately recognize faces has
significant applications in various fields, such as
security and surveillance. Further research is needed to
optimize the AI's performance in different scenarios,
ensuring its ethical and responsible use.
9.

In conclusion, our experiment demonstrates that AI is

capable of accurately recognizing faces in a variety of
lighting conditions and poses, with an average accuracy
of 90%. Our analysis of the data reveals that the AI is
particularly effective in recognizing faces with clear
and well-defined features while struggling with images
where the faces are partially obstructed or poorly lit.
Despite these limitations, the results of our study
suggest that AI has enormous potential in applications
such as security, surveillance, and identification. Future
research should focus on refining the AI algorithms and
improving the training data to further enhance its
accuracy and reliability.

11.

Confusion Matrix
AI Trained on 10 Drawings per Class
Classified as “Happy” Classified as “Sad”
Happy nnnn9nnnn nnnn1nnnn
Face
Sad Face nnnn1nnnn nnnn9nnnn
12.