Abstract:Real-time object detection is a very challenging task, as it requires both high accuracy and high speed. One-stage object detectors such as YOLO models are very fast but they are also less accurate than two-stage object detectors such as Faster R-CNN. However, Faster R-CNN is not as fast as the YOLO models. In this study, we propose an ensemble approach to real-time object detection that combines the strengths of YOLOv5 and Faster R-CNN. We ﬁrst use YOLOv5 to quickly generate a set of object proposals. We then use Faster R-CNN to reﬁne these proposals and produce more accurate object detection results. To further improve the accuracy of our object detection results, we propose a cascade reﬁnement network that uses dynamic ﬁne-tuning. The cascade reﬁnement network uses Kullback-Leibler divergence to dynamically adjust the weights of the Faster R-CNN model based on the conﬁdence scores of the YOLOv5 object proposals. We evaluated our proposed approach on the novel dataset collected in Uganda with other State-of-the-art approaches which include RetinaNet, Cascade R-CNN, Single-Shot MultiBox Detector (SSD), and Region-based Convolutional Neural Network (R-CNN). Experimental results revealed that the proposed ensemble model outperformed both base models with an average precision of 0.96, which is signiﬁcantly higher than the average precision of 0.91 for YOLOv5 and 0.90 for Faster R-CNN. The ensemble model was also able to achieve real-time inference speeds, processing frames at a rate of 25 frames per second, the same speed achieved by YOLOv5, faster than the speed of 15 frames per second by Faster R-CNN. The results also revealed that the proposed ensemble model is comparable to other state-of-the-art object detection models. Our proposed approach can be used to improve the accuracy and speed of real-time object detection in a variety of applications.

Chapter 8 Object Detection

On the Design of Cascades of Boosted Ensembles for Face Detection

Training a Multi-Exit Cascade with Linear Asymmetric Classification for Efficient Object Detection

Efficiently Learning a Detection Cascade with Sparse Eigenvectors

Asymmetric Totally-Corrective Boosting for Real-Time Object Detection

Boosting 2-Thresholded Weak Classifiers over Scattered Rectangle Features for Object Detection.

Fast Asymmetric Learning for Cascade Face Detection

Cascade Object Detection with Complementary Features and Algorithms

Fast and Robust Object Detection Using Asymmetric Totally Corrective Boosting

Real-Time Object Detection using an Ensemble of One Stage and Two Stage Object Detection Models with Dynamic Fine-tuning using Kullback-Leibler Divergence

A Trainable System for Object Detection

Recent advances in deep learning for object detection

A Survey of Deep Learning-based Object Detection

Multi-Object Detection Based On Deep Learning In Real Classrooms

Introduction to Computer Vision and Real Time Deep Learning-based Object Detection

Real-Time Object Detection and Recognition with Computer Vision

Cascade R-CNN: High Quality Object Detection and Instance Segmentation

Cascade R-CNN: Delving into High Quality Object Detection

Deep Learning for Generic Object Detection: A Survey