Authors :
Abhinav R. Diwakar; Kartik B. Kuri; Monika; Nagamani Grandhe; Bharani Kumar Depuru
Volume/Issue :
Volume 8 - 2023, Issue 12 - December
Google Scholar :
http://tinyurl.com/yv8eusmb
Scribd :
http://tinyurl.com/dyh7p777
DOI :
https://doi.org/10.5281/zenodo.10461395
Abstract :
Object detection has revolutionized industries
like manufacturing, healthcare, and transportation by
enabling automated object identification and
classification in images and videos. This paper explores
cutting-edge architectures like YOLOv3, YOLOv4 and
YOLOv8, highlighting their remarkable strides in
accuracy, speed, and robustness. YOLOv8, with its
balanced performance and ease of implementation, has
emerged as a leading architecture. However, practical
challenges like overfitting, data collection difficulties,
model complexity, and high hardware demands can
hinder its real-world adoption.
This paper investigates prominent toolboxes like
MMDetection and Detectron2 to address these
challenges. Detectron2 provides optimization techniques
and hardware acceleration strategies to tackle model
complexity and hardware demands.
The paper also explores deploying YOLOv8 on
Streamlit/Flask platforms, offering a user-friendly
interface for interacting with the model and visualizing
its detections, facilitating integration into web
applications. Strategies for overcoming deployment
challenges and achieving optimal performance are also
discussed.
Looking ahead, the paper investigates data-driven
pipe detection and classification for enhanced inventory
management. This promising approach utilizes
computer vision algorithms to automate pipe
identification and categorization, potentially
revolutionizing inventory management practices and
streamlining operations.
By addressing YOLOv8 implementation challenges
and exploring promising future directions, this paper
contributes to the advancement of object detection
technologies and their transformative impact across
various industries.
Keywords :
Object Detection, YOLOv8, Pipe Inventory Management, Image Processing, Artificial Intelligence, Computer Vision, Deep Learning, Object Tracking, Real- time Object Counting, Automated Inventory Management, Pipe Classification.
Object detection has revolutionized industries
like manufacturing, healthcare, and transportation by
enabling automated object identification and
classification in images and videos. This paper explores
cutting-edge architectures like YOLOv3, YOLOv4 and
YOLOv8, highlighting their remarkable strides in
accuracy, speed, and robustness. YOLOv8, with its
balanced performance and ease of implementation, has
emerged as a leading architecture. However, practical
challenges like overfitting, data collection difficulties,
model complexity, and high hardware demands can
hinder its real-world adoption.
This paper investigates prominent toolboxes like
MMDetection and Detectron2 to address these
challenges. Detectron2 provides optimization techniques
and hardware acceleration strategies to tackle model
complexity and hardware demands.
The paper also explores deploying YOLOv8 on
Streamlit/Flask platforms, offering a user-friendly
interface for interacting with the model and visualizing
its detections, facilitating integration into web
applications. Strategies for overcoming deployment
challenges and achieving optimal performance are also
discussed.
Looking ahead, the paper investigates data-driven
pipe detection and classification for enhanced inventory
management. This promising approach utilizes
computer vision algorithms to automate pipe
identification and categorization, potentially
revolutionizing inventory management practices and
streamlining operations.
By addressing YOLOv8 implementation challenges
and exploring promising future directions, this paper
contributes to the advancement of object detection
technologies and their transformative impact across
various industries.
Keywords :
Object Detection, YOLOv8, Pipe Inventory Management, Image Processing, Artificial Intelligence, Computer Vision, Deep Learning, Object Tracking, Real- time Object Counting, Automated Inventory Management, Pipe Classification.
