Vehicle Detection from Aerial Images Using Deep Learning: A Comparative Study

Abstract

This paper addresses the problem of car detection from aerial images using Convolutional Neural Networks (CNNs). This problem presents additional challenges as compared to car (or any object) detection from ground images because the features of vehicles from aerial images are more difficult to discern. To investigate this issue, we assess the performance of three state-of-the-art CNN algorithms, namely Faster R-CNN, which is the most popular region-based algorithm, as well as YOLOv3 and YOLOv4, which are known to be the fastest detection algorithms. We analyze two datasets with different characteristics to check the impact of various factors, such as the UAV’s (unmanned aerial vehicle) altitude, camera resolution, and object size. A total of 52 training experiments were conducted to account for the effect of different hyperparameter values. The objective of this work is to conduct the most robust and exhaustive comparison between these three cutting-edge algorithms on the specific domain of aerial images. By using a variety of metrics, we show that the difference between YOLOv4 and YOLOv3 on the two datasets is statistically insignificant in terms of Average Precision (AP) (contrary to what was obtained on the COCO dataset). However, both of them yield markedly better performance than Faster R-CNN in most configurations. The only exception is that both of them exhibit a lower recall when object sizes and scales in the testing dataset differ largely from those in the training dataset.

Vehicle Detection for Vision-Based Intelligent Transportation Systems Using Convolutional Neural Network Algorithm

“…In study proposed by Alawi et al, authors present the problems facing vehicle detection systems from aerial images using neural networks such as Faster R-CNN. It is sometimes difficult for the comparison between vehicles and objects to distinguish between them [17,18]. e researchers studied the capabilities of a neural network algorithm in addition to YOLOv3, YOLOv4, and their performance in detection application.…”

Deep-Learning-Based Automated Palm Tree Counting and Geolocation in Large Farms from Aerial Geotagged Images

“…In this section, we present a brief overview of the four state-of-the-art object detectors that we tested, as well as the mathematical background of the technique that we applied for geolocating the detected objects. The selected models are representative of the recent trends in the families of one-stage and two-stage object detectors and have been proven successful in terms of average precision and inference speed in a wide variety of applications [20][21][22][23][24][25].…”

Deep Learning for UAV-based Object Detection and Tracking: A Survey

“…However, less attention has been paid to the advance of object and tracking techniques both in image and video acquired by UAV. Although reviews in [39]- [44] present some DL-based static object detection for UAV image and the one in [45] presents traditional object tracking for UAV video, there still lacks a complete survey for object and tracking and the most recent advances.…”

“…[78], [87] modified the feature resolution of the lightweight Pelee network [111] to meet real-time needs. Due to the efficiency and power of YOLOv4, many object detection models [44], [112] are based on this network. Ammar et al [44] used YOLOv3 and newly released YOLOv4 to detect vehicles with inference processing speed from 12 fps for 608 × 608 up to 23 fps for 320 × 320.…”

“…Due to the efficiency and power of YOLOv4, many object detection models [44], [112] are based on this network. Ammar et al [44] used YOLOv3 and newly released YOLOv4 to detect vehicles with inference processing speed from 12 fps for 608 × 608 up to 23 fps for 320 × 320. Furthermore, Wang et al [113] designed a Strip Bottleneck with YOLO network (SPB-YOLO) based on YOLOv5 for engineering application.…”