Bioinspired Garra Rufa Optimization-Assisted Deep Learning Model for Object Classification on Pedestrian Walkways.

Area of Science:

• Computer Vision
• Deep Learning
• Artificial Intelligence

Background:

• Object detection in pedestrian areas is vital for safety but manual labeling of abnormal actions is tedious.
• Advancements in deep learning (DL) offer potential for automated surveillance systems.
• Computer vision (CV) researchers require efficient methods for object detection and classification.

Purpose of the Study:

• To design a bioinspired Garra rufa optimization-assisted deep learning model for object classification (BGRODL-OC) on pedestrian walkways.
• To recognize the presence of pedestrians and objects in surveillance videos.
• To enhance automatic anomaly identification in CV.

Main Methods:

• Utilized GhostNet feature extractors for generating feature vectors.
• Employed the Garra rufa optimization (GRO) algorithm for hyperparameter tuning.
• Implemented an attention-based long short-term memory (ALSTM) network for object classification.

Main Results:

• The BGRODL-OC technique demonstrated superior performance in object classification on pedestrian walkways.
• Experimental analysis validated the effectiveness of the proposed method.
• The BGRODL-OC algorithm outperformed existing approaches in detecting pedestrians and objects.

Conclusions:

• The BGRODL-OC technique provides an effective solution for automatic object detection and classification in pedestrian surveillance.
• This approach significantly improves the efficiency and accuracy of anomaly detection systems.
• The study highlights the potential of bioinspired optimization combined with deep learning for enhancing CV applications in safety-critical domains.