Design and implementation of real-time object detection system based on single-shoot detector and OpenCV

Area of Science:

• Computer Vision
• Human-Computer Interaction

Background:

• Real-time object detection is crucial for applications like inspection systems.
• Existing methods require further optimization for speed and accuracy on accessible hardware.

Purpose of the Study:

• To develop and implement an efficient real-time object detection and recognition system.
• To evaluate pre-trained deep learning models with the Single-Shot Detector (SSD) algorithm for accuracy and speed.
• To ensure the system operates effectively on standard computing equipment.

Main Methods:

• Utilized deep learning techniques and the Single-Shot Detector (SSD) algorithm.
• Employed pre-trained models evaluated on diverse datasets (MS COCO, PASCAL VOC, Kitti).
• Developed and tested various deep learning architectures for optimal performance.

Main Results:

• Achieved a high accuracy of 97% in real-time object detection and recognition.
• Demonstrated the system's capability to detect both static and moving objects.
• Identified optimal pre-trained models for the SSD algorithm regarding accuracy and speed.

Conclusions:

• The proposed system offers a highly accurate and efficient solution for real-time object detection.
• The integration of deep learning with SSD provides a robust framework for object recognition tasks.
• The system is suitable for deployment on reasonably equipped hardware, broadening its applicability.