ABSTRACT

Autonomous driving has emerged as one of the most transformative

technologies of the 21st century, relying heavily on the seamless integration
of computer vision, deep learning, embedded systems, and sensor fusion.
With the growing complexity of real-world environments and the demand for
efficient and reliable decision-making, multiple research directions have
emerged to enhance the perception, adaptability, and computational
performance of autonomous systems. This abstract provides a comprehensive
synthesis of six significant contributions that collectively advance the state of
the art in pedestrian detection, weather-robust perception, embedded AI, and
multimodal sensor fusion for autonomous vehicles. One of the foundational
challenges in deploying vision-based systems in autonomous vehicles is the
trade-off between performance and hardware constraints. The study
titled*"Towards Reconfigurable CNN Accelerator for FPGA
Implementation"* introduces a CNN accelerator architecture that is capable
of reconfiguration during runtime. Two architectural models are compared: a
fused model and a branched model. Both exploit layer-wise sharing across
multiple tasks to optimize performance and minimize resource overhead. This
reconfigurable design significantly benefits real-time inference on embedded
FPGA platforms, enabling adaptive operation in diverse driving conditions
with limited computational budgets. In adverse weather conditions such as
fog, rain, and low light, conventional object detection systems often degrade
in performance.
 ABSTRACT

This abstract presents a comprehensive overview of fake news detection

methodologies, categorizing them into content-based and context-based
approaches. Content-based techniques focus on analyzing the linguistic and
semantic features of news articles using NLP tools. These include lexical and
syntactic patterns, sentiment analysis, readability scores, and stylistic cues that
may reveal deception. Traditional machine learning classifiers—such as Support
Vector Machines (SVM), Naive Bayes, and logistic regression—have been
widely used to train models on labeled datasets. More recently, deep learning
models, including convolutional neural networks (CNNs), recurrent neural
networks (RNNs), and transformer-based architectures such as BERT, have
demonstrated superior performance in capturing contextual and semantic nuances
in textual data. Context-based approaches go beyond textual content and
incorporate external information such as the credibility of the news source, the
behavior of users interacting with the news (e.g., sharing and commenting), and
the way the news propagates across social media networks. Graph-based models
and user-news interaction networks are increasingly used to analyze the spread
and credibility signals. Moreover, real-time data from fact-checking platforms
and knowledge bases can be integrated to verify claims more effectively. Despite
technological advancements, the field faces several ongoing challenges. These
include the dynamic and adaptive nature of fake news, scarcity of high-quality
labeled datasets, cross-domain generalization issues, and the black-box nature of
many machine learning models. There are also ethical considerations, such as the
potential for censorship, the need to protect freedom of expression, and the
importance of model transparency and explain ability. In conclusion, fake news
detection is a multifaceted problem that demands a combination of robust
computational models, linguistic insight, and ethical governance.
 ABSTRACT

Medical imaging has become a cornerstone in healthcare diagnostics,

especially in detecting and analyzing internal abnormalities without invasive
procedures. Challenges such as low image quality, noise, and complex
anatomical variations necessitate robust computational techniques for accurate
interpretation. Advanced methods such as Convolutional Neural Networks
(CNNs) and hybrid deep learning models are widely adopted to improve
diagnostic reliability and automate tasks like segmentation, classification, and
anomaly detection. A common approach integrates deep neural networks with
classical image enhancement and feature extraction pipelines, which increases
precision in interpreting MRI, CT, and X-ray images. These hybrid strategies
help overcome limited dataset sizes by employing transfer learning and data
augmentation, enhancing generalization across clinical scenarios. Additionally,
multimodal imaging analysis combines different scan types (e.g., PET-CT, MRI-
CT) and patient metadata to provide comprehensive insights. This fusion
significantly enhances diagnostic power, especially in complex fields like
oncology and neurology. In conclusion, the synergy between artificial
intelligence and imaging technology is driving significant advancements in
medical diagnostics, enabling earlier detection, personalized treatment planning,
and improved patient outcomes.
VIVEKANANDHA COLLEGE OF ENGINEERING FOR WOMEN
(Autonomous)
Elayampalayam, Tiruchengode, Namakkal Dt.- 637205

III YEAR

BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE AND TECHNOLOGY

FAKE NEWS DETECTION

RESEARCH PAPERS

Submitted by

[Link] STUDENT NAME REGISTER NUMBER

1. MONIKA M 612922107031

Under the guidance of

Mrs. S. FOWJIYA M.E.,

ASSISTANT PROFESSOR/CST
VIVEKANANDHA COLLEGE OF ENGINEERING FOR WOMEN
(Autonomous)
Elayampalayam, Tiruchengode, Namakkal Dt.- 637205

III YEAR

BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE AND TECHNOLOGY

MEDICAL IMAGE ANALYSIS

RESEARCH PAPERS

Submitted by

[Link] STUDENT NAME REGISTER NUMBER

1. SRILAKSHMIMALINI V 612922107052

Under the guidance of

Mrs. S. FOWJIYA M.E.,

ASSISTANT PROFESSOR/CST
 ABSTRACT

Animal-vehicle collisions (AVCs) pose a serious threat to both human life

and wildlife populations, particularly in areas where roadways intersect with natural
habitats. The integration of technology into wildlife conservation has led to
innovative solutions aimed at mitigating these incidents. This study explores the
development and implementation of animal detection systems near roads using a
combination of computer vision, sensor networks, and artificial intelligence. The
primary goal is to enhance driver awareness in real-time and reduce the frequency
and severity of AVCs. Various detection methods, such as infrared sensors, motion
detectors, thermal cameras, and deep learning-based image recognition, are analyzed
for their accuracy, reliability, and cost-effectiveness .Moreover, this research
evaluates the integration of these technologies into existing infrastructure,
considering factors like power supply, and data transmission. By employing deep
learning models such as convolutional neural networks (CNNs), the detection
accuracy of animals in diverse lighting and environmental conditions has shown
significant improvement. Field experiments conducted in wildlife-prone areas
demonstrate a notable decrease in AVC occurrences post-deployment of such
systems . The abstract also addresses the challenges associated with large-scale
implementation, including the cost of deployment, maintenance, and the necessity
for cross-sector collaboration between transportation departments, wildlife
conservation agencies, and technology providers. Policy implications and future
research directions are discussed, adaptive solutions that can respond to changing
wildlife behaviors and road use patterns. In conclusion, animal detection systems
near roads present a promising avenue to ensure both road safety and wildlife
conservation, paving the way for smarter and safer transportation infrastructure.
VIVEKANANDHA COLLEGE OF ENGINEERING FOR WOMEN
(Autonomous)
Elayampalayam, Tiruchengode, Namakkal Dt.- 637205

III YEAR

BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE AND TECHNOLOGY

ENHANCING ROAD SAFETY WITH

AUTOMATED ANIMAL DETECTION

RESEARCH PAPERS

Submitted by

[Link] STUDENT NAME REGISTER NUMBER

1. MARUSHINI B 612922107030

Under the guidance of

Mrs. S. FOWJIYA M.E.,

ASSISTANT PROFESSOR/CST
VIVEKANANDHA COLLEGE OF ENGINEERING FOR WOMEN
(Autonomous)
Elayampalayam, Tiruchengode, Namakkal Dt.- 637205

III YEAR

BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE AND TECHNOLOGY

SELF DRIVING CARS

RESEARCH PAPERS

Submitted by

[Link] STUDENT NAME REGISTER NUMBER

1. HARINI M 612922107017

Under the guidance of

Mrs. S. FOWJIYA M.E.,

ASSISTANT PROFESSOR/CST