Multi-Task Cost-Sensitive-Convolutional Neural Network for Car Detection

Abstract

This paper proposes a novel smart parking scheme for the parking lot. Automatic car detection is the core technology of the proposed scheme. However, new challenges arise in car detection in aerial views, such as a large number of tiny objects and complex backgrounds. In order to solve these issues, this paper proposes a car detection method based on multi-task cost-sensitive-convolutional neural network (MTCS-CNN). In the proposed network framework, multi-task partition layer which is composed of some sub-task selection units is first developed. The sub-task selection unit is constructed by introducing local mask and non-zero pooling, which can divide the complex detection task into many simple sub-tasks. To tackle the obtained sub-tasks, cost-sensitive sub-network is proposed based on faster R-CNN framework with the introduction of cost-sensitive loss. In the proposed Multi-task partition layer, the sub-task selection unit is used to capture the local map of the original aerial view image. In each local map, the scale and the number of objects are enlarged and decreased, respectively. Therefore, multi-task partition layer can divide a complex tiny objects detection task into many simple enlarged objects detection sub-tasks, which is helpful for performance improvement. In addition, the proposed cost-sensitive loss can effectively discount the effect of easy examples and focus attention on the hard examples, which may improve the detection performance on hard examples. Therefore, the model with incorporation of proposed cost-sensitive loss is robust to the complex background, further improving the performance. On our dataset, the proposed method obtained an mAP of 85.3%, outperformed state-of-the-art method. INDEX TERMS Smart parking, car detection, multi-task cost-sensitive-convolutional neural network.

Optimized Convolutional Neural Network-Based Object Recognition for Humanoid Robot

“…In recent years, many CNN-based studies have focused on such problems as small object detection [6], face detection [20], crowd counting [21,22], traffic sign detection [23], and car detection [24]. In the meantime, there are increasing research on the deep learning application for the humanoid robot's object recognition, grasp detection, etc.…”

Bounding Box-Free Instance Segmentation Using Semi-Supervised Iterative Learning for Vehicle Detection

“…In-ground view images, several literature reviews address advanced driver assistance systems (ADAS) for autonomous vehicles using image processing and vehicle detection from various onboard handling sensors such as radar, monocular camera, and camera binocular [3], [4], [5]. In addition, several studies use images from surveillance cameras on roads [6], on top of buildings [7], pedestrian bridges [8], among others.…”

DeepBrain: Experimental Evaluation of Cloud-Based Computation Offloading and Edge Computing in the Internet-of-Drones for Deep Learning Applications

“…While the main objective of drones is to collect visual data including aerial images and videos (in addition to other types of data), deep learning algorithms are nowadays the de facto standard for processing images and extracting useful data using different techniques such as classification, object detection, semantic segmentation, and instance segmentation [ 4 ]. Several recent research works have leveraged deep learning algorithms based on convolutional neural networks to process aerial images collected from drones [ 5 , 6 , 7 , 8 , 9 ]. In [ 5 , 6 ], the authors conducted a comparative study between the two state-of-the-art algorithms YOLOv3 and Faster Region-CNN (RCNN) for the detection of cars from aerial images.…”

“…Ammour et al in [ 8 ] developed a pre-trained CNN in addition to a support vector machine (SVM) classifier for the detection and counting of vehicles from high-resolution drone images. In [ 9 ], the authors investigated the problem of vehicle tracking from aerial images. The authors in [ 10 ] proposed CNN architectures to automate the classification of aerial scenes of disaster events, such as fires, earthquakes, floods, and accidents.…”