Practical No: 1

AIM: Literature Review On the state-of-the-art Image

Processing Applications.

Project Title AI-Powered Home Control System

for Elderly People
UsingComputer Vision
Team Itisha Sakhreliya (21BT04108)
Members
Mansi Surti(21BT04127)
Literature
Survey:
Research Paper Brief Summary Techniques used Libraries/Tools used Dataset Used Pros and
Cons
1. The proposed framework aims to Compter OpenCV UR Fall Pros:
Elderly address the common challenge of VisionDeep TensorFlow Detection Enhanced
Fall Detection visibility obstruction in real-world Learning OpenPose Dataset Accuracy, Real-
Using CCTV
environments, ensuring that fall Pose NVIDIA GPUs NTU RGB+D TimeProcessing,
Cameras
detection remains reliable even when Estimation Dataset: Scalability.
under Partial
parts of the body are not visible to Short-Time Fourier
Occlusion of
the Subject's the Transform(STFT) Cons:
Body. camera. Complexity,
Dependency on
Visual Data,
Resource Intensive.

GSFCU/SOT/CSE/ BTCS701E/21BT04127 1
2. An This paper introduces a modular Convolutional Neural TensorFlow/Keras Multiple Pros:
Interpretable deep Networks Cameras
Modular
Deep Learning
learning framework (CNNs) OpenPose Fall Dataset High Accuracy and
specifically Efficiency,
Framework designed
for for real-time fall Short-Time Fourier OpenCV (MCFD) Interpretability, Modular
detection Transform
Video-Based usingFall
video data. The system focuses (STFT) NVIDIA Tesla V100 UR Fall Design.
on Detection
Detection providing high accuracy and Pose Estimation GPU Dataset Cons:
efficiency, Gradient-weighted
NTU RGB+D Complex System
ClassActivation
Which is critical for application Mapping (Grad andNTU Architecture,
inelderly care. CAM)
RGB+D Dependency on
120
High-Performance

Hardware,
Scalability
Challe
nges in
Diverse
Environments.

GSFCU/SOT/CSE/ BTCS701E/21BT04127 2
3. A Survey on The paper categorizes Traditional Computer Vision OpenCV Multiple Cameras Pros:
Vision-Based Fall and analyzes various Techniques TensorFlow/PyTorch Fall Dataset Comprehensive Coverage,
Detection: Current methods, from Machine Learning and OpenPose (MCFD) Identification of Challenges,
Trends and Future traditional computer DeepLearning Le2i Fall Informed Insights.
Challenges vision approaches to Pose Estimation and Detection Dataset
more advanced deep HumanBody Modeling NTU RGB+D Cons:
learning-based Multimodal Data Fusion Dataset Broad Scope, Focus on
techniques, also Vision-Based
discusses future Method
directions for improving s,
the robustness, Rapidly Evolving Field.
accuracy, and
practicality of these
systems in real-world
environments.

GSFCU/SOT/CSE/ BTCS701E/21BT04127 3
4. Deep Learning- It provides an in-depth Convolutional Neural TensorFlow/Keras Multiple Cameras Pros:
Based Fall analysis of various deep Networks(CNNs) OpenCV Fall Dataset Comprehensive
Detection learning models, Recurrent Neural PyTorch (MCFD) Coverag
System for highlighting their Networks(RNNs) Le2i Fall e,Identification of
Elderly Care strengths and Transfer Detection Dataset Challenges.
Using weaknesses in fall Learning Data NTU RGB+D
Surveillance detection tasks. The Augmentation Dataset Cons:
Cameras research discusses the Broad Scope, Focus on
potential of these Vision-Based
models to achieve high Method
accuracy and reliability s,
in real-time Rapidly Evolving Field.
applications.

GSFCU/SOT/CSE/ BTCS701E/21BT04127 4
5. Real-Time Fall The study focuses on Background OpenCV Le2i Fall Pros:
Detection for achieving a balance SubtractionOptical MATLABCUD Detection Dataset Real-Time
Elderly Using between detection Flow Weizmann Human Performanc
Computer accuracy and Histogram of Oriented Action Dataset e,Computational
Vision computational Gradients(HOG) Efficiency,
efficiency, making the Support Vector Machine Non-Invasive Monitoring.
(SVM)
system suitable for Cons:
practical deployment. Environmental Sensitivity,
By leveraging False Positives/Negatives,
advanced ResourceDependency.
image processing and
motion analysis
methods, the system can
accurately detect falls
in real-time, providing
immediate alerts to
caregivers.

GSFCU/SOT/CSE/ BTCS701E/21BT04127 5