Faster and Real-Time Object Detection System using YOLOv3-tiny in Comparison with Mobilenet SSD Network

The proposed work aims to detect numerous objects from various classes effectively using YOLOv3-tiny in comparison with mobilenet SSD algorithm. Materials and Methods: A sample size of 91images were gathered for this research by determining considering the power of components dimension of the effect and average error rate which were 0.21, 0.05, and 0.80. The COCO dataset is used to construct a dataset for real-time detection. The configuration and weights files for YOLOv3-tiny are made up of several variables, architectures, and measures for the objects which were available in the dataset (g power = 0.8). Results: The proposed approach detects items at a faster rate of 220 FPS and with a higher accuracy of 70%, whereas the SSD network recognises several objects at a time at a faster rate of 46 FPS and with precision of 85% (p.001).