Real-time object detection is a difficult task
that has drawn a lot of interest in the deep learning
community. Object detection algorithms are frequently
employed in robotics, security, and autonomous car
applications. In this abstract, we suggest a novel deep
learning method for real-time object detection. You Only
Look Once (YOLO) and Faster R-CNN (Region-based
Convolutional Neural Network), two well-known deep
learning architectures, are the foundation of our
suggested solution. The Faster R-CNN design is
renowned for its accurate object localisation, whereas
the YOLO architecture is noted for its speed and
accuracy in object recognition. In order to quickly locate
potential object regions in the input image, we suggest
using the YOLO architecture. After that, a Faster RCNN network is used to accurately localise the items
within these candidate regions. We can perform realtime object detection with high accuracy and exact
localisation by fusing the benefits of these two systems.
We offer a novel loss function that combines the YOLO
and Faster R-CNN loss functions in order to
substantially boost the performance of our method. With
the use of this loss function, we can train our network to
simultaneously optimise for speed and accuracy, creating
a more effective system for object detection. Our
suggested method has been rigorously tested on
numerous datasets, and the findings demonstrate that it
performs better in terms of speed and accuracy than
cutting-edge object detection algorithms. We think that
our strategy has the potential to revolutionise the realtime object identification industry and open the door for
the creation of fresh, cutting-edge applications.
Keywords :
object detection, deep learning, cnn