PeleeNet Algorithm Optimization based on Feature Pyramid Networks for Object Detection

Abstract

Faced with the increasing demand for image recognition on mobile devices, how to run convolutional neural network (CNN) models on mobile devices with limited computing power and limited storage resources encourages people to study efficient model design. In recent years, many effective architectures have been proposed, such as mobilenet_v1, mobilenet_v2 and PeleeNet. However, in the process of feature selection, all these models neglect some information of shallow features, which reduces the capture of shallow feature location and semantics. For multi-scale target detection model, FPN (feature pyramid networks) network has a good detection effect. Most of the target detection algorithms only use deep features to predict, but we know that the shallow feature semantic information is relatively small, but the target location is accurate. The deep feature semantic information is rich, but the target location is rough. We optimize the original PeleeNet model and FPN model and use bilinear interpolation algorithm in FPN model to enlarge the feature map. This algorithm makes full use of the four real pixel values around the virtual point in the source image to determine a pixel value in the target image, so the magnification effect is much better than the simple nearest neighbor interpolation. In this paper, we propose an effective framework based on Feature Pyramid Networks to improve the recognition accuracy of deep and shallow features while guaranteeing the recognition speed of PeleeNet structured images. Compared with PeleeNet, the structure recognition accuracy of cifar-10 data set is improved by 4.0% and the recognition speed is increased by 31.4%.