International Journal of Computer Applications
Foundation of Computer Science (FCS), NY, USA
|
Volume 186 - Issue 74 |
Published: March 2025 |
Authors: Pranta Kumar Sarkar, Moskura Hoque, Mostofa Kamal Nasir |
![]() |
Pranta Kumar Sarkar, Moskura Hoque, Mostofa Kamal Nasir . Vision-Based Human Activity Recognition Uses A Deep Learning Approach. International Journal of Computer Applications. 186, 74 (March 2025), 70-74. DOI=10.5120/ijca2025924621
@article{ 10.5120/ijca2025924621, author = { Pranta Kumar Sarkar,Moskura Hoque,Mostofa Kamal Nasir }, title = { Vision-Based Human Activity Recognition Uses A Deep Learning Approach }, journal = { International Journal of Computer Applications }, year = { 2025 }, volume = { 186 }, number = { 74 }, pages = { 70-74 }, doi = { 10.5120/ijca2025924621 }, publisher = { Foundation of Computer Science (FCS), NY, USA } }
%0 Journal Article %D 2025 %A Pranta Kumar Sarkar %A Moskura Hoque %A Mostofa Kamal Nasir %T Vision-Based Human Activity Recognition Uses A Deep Learning Approach%T %J International Journal of Computer Applications %V 186 %N 74 %P 70-74 %R 10.5120/ijca2025924621 %I Foundation of Computer Science (FCS), NY, USA
In today's world, daily life increasingly depends on vision-based advanced technologies, which enhance the reliability and convenience of human lifestyles. Among these technologies, vision-based Human Activity Recognition (HAR) stands out as a comprehensive and challenging field of study, with broad exploration and practical applications. HAR systems are designed to identify diverse human actions under varying environmental conditions.Vision-based activity recognition plays a crucial role in a wide range of applications, including user interface design, robot learning, security surveillance, healthcare, video searching, abnormal activity detection, and human-computer interaction. This study focuses on recognizing various human activities in real-world settings, highlighting the importance of consistency and credibility in the results.To achieve this, data was collected from multiple sources and processed using three distinct models—Convolutional Neural Network (CNN), VGG-16, and ResNet50—to identify the most effective approach for activity recognition. Among these, a specific architectural CNN model was further evaluated for its ability to capture human activity features in specific video sequences. The training, validation, and testing phases utilized a comprehensive dataset comprising 56,690 images. Remarkably, the proposed system achieved an impressive accuracy of 96.23% after 30 epoch running and low validation loss illustrate its effectively recognition each feature.