TECHNICAL SESSION

Generative AI And Education:

By: Prof. Rafael Ferreira Mello
Professor Rafael Ferreira Mello talked about how generative AI, like ChatGPT, is changing
education. He investigated the impact of these tools on content creation,
personalization, and consumption within educational settings. Mello emphasized the
advantages of enhanced learning experiences and personalized content delivery, while
also acknowledging obstacles such as academic honesty and ethical use. He stressed the
significance of responsibly incorporating AI to enhance, rather than substitute,
conventional teaching approaches. Furthermore, Mello discussed examples of projects
he worked on in Brazil, such as Duolingo, Story Spark, and Tutoria.

Urban Pluvial Flood Prediction Using Image Processing:

By: Sajid Majeed
Sajid Majeed's research project is centered on the application of machine learning (ML)
to rapidly and precisely anticipate urban flooding triggered by intense rainfall. The
model uses rainfall intensity and urban infrastructure analysis to identify patterns for
quicker flood predictions compared to traditional methods, making it suitable for real-
time alerts and early warning systems. The research emphasizes the worldwide
consequences of floods, especially in South Asia, and the significance of using geo-
spatial methods for efficient flood control. Majeed trained a machine learning model on
rainfall data from six Sindh cities in Pakistan for 65 years. The model used a 140.9 mm
rainfall threshold to categorize years as either "flood" or "no flood." Out of four ML models
examined, Logistic Regression had the highest accuracy rate at 97%. The study highlights
the importance of using machine learning in predicting floods and recommends
exploring more weather factors to enhance forecasting accuracy.

Early Stage Breast Cancer Prediction Using Machine Learning

Models:
By: Maria Munawwar
Maria Munawwar's project is centered on improving the detection of breast cancer by
utilizing advanced image processing and artificial intelligence methods. Her goal is to
enhance mammogram image quality through Wiener filtering and utilize Vision
Transformers (ViT) in machine learning to detect cancer sooner, which is essential for
successful treatment. The project utilizes the CBIS-DDSM dataset to categorize breast
abnormalities into four groups: benign mass, benign calcification, malignant mass, and
malignant calcification. After data processing and augmentation, a ResNet-50 model
achieved an accuracy of 81%. The study also includes a React JS user-friendly interface
that simplifies the task of doctors in handling patient information and creating reports.
Even though there are obstacles such as managing big image files and limited hardware
capabilities, the research shows the system's promise for use in clinical settings and
proposes enhancements in accuracy of predictions and processing capabilities.
Real Time Detection Of Cyber Threats Via Dark Web Traffic Analysis
Using Machine Learning And Deep Learning:
By: Aveen Faheem
This study delves into enhancing cybersecurity by examining dark web traffic in order to
identify cyber threats, with a specific emphasis on pinpointing common attack methods
and averting data breaches involving private personal data. The research utilizes the
CICDarknet 2020 dataset containing benign and dark web traffic, including audio, emails,
and videos. Different machine learning algorithms, such as Decision Tree, Random Forest,
XGBoost, and Neural Networks, are used for identifying unusual occurrences in network
traffic. The XGBoost model notably achieved a 98% accuracy rate in detecting cyber
threats. The suggested model can be used to examine dark web traffic for improving
cybersecurity. Future research will focus on investigating AI techniques for improving
packet detection efficiency. Aveen Faheem's talk emphasizes the utilization of machine
learning and deep learning to detect cyber threats in real-time dark web traffic, stressing
the importance of focused analysis in identifying attack patterns overlooked by
traditional security systems.

Safecampus: Leveraging AI For Advanced Surveillance And Security

Enhancement:
By: Hamna Farooq
Hamna Farooq's initiative, Safe Campus, aims to enhance campus safety through AI-
driven surveillance. The system uses object and facial recognition technologies to
analyze live CCTV footage for identifying threats such as suspicious behavior,
trespassing, and unauthorized access. It is connected to the current campus
infrastructure and offers proactive security by notifying on-duty staff as soon as a threat
is identified, allowing for quicker, data-informed decision-making. The AI system includes
cutting-edge video analytics, like the YOLOv8 model, to identify particular incidents such
as no-parking violations and large crowds. Even though there are difficulties in detecting
uncommon and tiny objects, the system manages to reach an 80.56% accuracy rate in
detection and is constantly enhancing its performance with additional data and model
adjustments. The project improves campus safety by allowing faster response times and
creating a more secure atmosphere.

Recommender System For Educational And Professional Growth:

By: Summiya Iqbal Jangda
Summiya Iqbal Jangda is dedicated to enhancing personalized recommendation systems
for educational resources, career paths, and skill-building opportunities through her
research. Through the utilization of machine learning, these platforms recommend
personalized courses and resources according to users' preferences and previous
achievements. One major challenge that is tackled is the issue of "cold start," in which new
users with minimal data face difficulty in getting precise recommendations. The
suggested hybrid approach merges collaborative filtering and content-based filtering to
address this problem. Moreover, employing advanced deep learning methods like
LSTM/GRU layers and attention mechanisms improves the precision of the suggestions.
Future tasks will involve integrating deep reinforcement learning, utilizing real-time data,
and broadening recommendations to encompass a wider range of educational
resources. The research is intended to help users make well-informed choices for
academic and professional advancement, much like receiving product recommendations
on platforms such as Amazon.

Blockchain: Sharing Model To Increase The Efficiency Of Waste

Management Processes Based on Blockchain Technology:
By: Dr. Inayatulloh
This study investigates the utilization of blockchain sharding to enhance waste
management systems through improved scalability and data management
capabilities. Sharding splits data into multiple sections to allow quicker and more
effective handling of large quantities of waste-related data. The aim is to
establish a waste management model that is more sustainable and efficient.

Sonicvision: Navigating Object Counting Frontiers With Ultrasonic

Economy:
The HC-SR04 sensor is used in the development of an ultrasonic object counting
system called "SonicVision," which includes system modeling, Proteus simulation,
and hardware implementation. It involves comparing the prototype with current
market solutions. The project also looks into potential future uses in automated
counting and other economical ultrasonic solutions, highlighting the system's
versatility in different industries.