1. A novel automated image and video object detection based on bio-inspired method seagull and whale optimization algorithm using deep learning approach.
- Author
-
Kuppan, K., Bhavani, R., Mukil, A., and Lakshamanan, R.
- Subjects
- *
METAHEURISTIC algorithms , *CONVOLUTIONAL neural networks , *HUMPBACK whale , *IMAGE recognition (Computer vision) , *MATHEMATICAL optimization , *OBJECT recognition (Computer vision) , *DEEP learning - Abstract
Improving object recognition in computer vision by combining the Whale Optimization Algorithm (WOA) with Seagull Optimization Algorithm (SOA), two nature-inspired metaheuristic optimization techniques. Object detection plays a critical role in automated object recognition and localization in images and videos. The ResNet architecture has established its effectiveness in learning deep representations, while convolutional neural networks (CNNs) have demonstrated exceptional object detection performance. Both the WOA and the Seagull Algorithm are well-known for their unique exploration and exploitation skills, inspired by natural actions. By leveraging the complementary capabilities of these algorithms, the proposed approach aims to enhance the accuracy and efficiency of object identification in deep learning models. The WOA specializes in global exploration and convergence, mimicking the hunting thoughts of humpback whales. Meanwhile, the Seagull Algorithm simulates the swarming behavior of seagulls, focusing on local search and exploitation. Extensive studies on popular object detection are carried out to validate the efficacy of the suggested technique benchmarks, including KITTI, COCO, CUSTOM and PASCAL VOC datasets. The results demonstrate that the fusion of the Seagull and Whale Optimization Algorithm with the ResNet model consistently outperforms traditional optimization techniques and baseline object detection methods. The algorithm exhibits improved accuracy, robustness, and generalization on various challenging scenarios, showcasing its potential for real-world applications. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF