Abstract
In recent years, almost all of the current top-performing object detection networks use CNN (convolutional neural networks) features. State-of-the-art object detection networks depend on CNN features. In this work, we add feature fusion in the object detection network to obtain a better CNN feature, which incorporates well deep, but semantic, and shallow, but high-resolution, CNN features, thus improving the performance of a small object. Also, the attention mechanism was applied to our object detection network, AF R-CNN (attention mechanism and convolution feature fusion based object detection), to enhance the impact of significant features and weaken background interference. Our AF R-CNN is a single end to end network. We choose the pre-trained network, VGG-16, to extract CNN features. Our detection network is trained on the dataset, PASCAL VOC 2007 and 2012. Empirical evaluation of the PASCAL VOC 2007 dataset demonstrates the effectiveness and improvement of our approach. Our AF R-CNN achieves an object detection accuracy of 75.9% on PASCAL VOC 2007, six points higher than Faster R-CNN.
Highlights
Object detection accuracy has been improved by using deep CNN (Convolutional Neural Network) nets [1]
We propose a new object detection network that fixes the disadvantages of feature fusion and the attention mechanism in the faster R-CNN in case of background interference and small target problems
The object detection network that we proposed in this article achieved an mAP of 75.9% on the dataset, PASCAL VOC 2007, 6 points higher than the Faster R-CNN because of feature fusion and the attention module
Summary
Object detection accuracy has been improved by using deep CNN (Convolutional Neural Network) nets [1]. Faster R-CNN is an end-to-end object detection network because it combines region proposal and detection into a unified network. The object detection networks have three essential components called the feature extraction net, region proposal network, and classification and regression. Object detection often uses the last CNN feature maps for region proposal net. Pixels are treated the same in the CNN feature maps and the region proposal net. We propose a new object detection network that fixes the disadvantages of feature fusion and the attention mechanism in the faster R-CNN in case of background interference and small target problems. The new CNN feature map combines fine, shallow layer information with coarse, deep layer information
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
