Convolutional Neural Networks for Aerial Vehicle Detection and Recognition

Soleimani, Amir; Nasrabadi, Nasser M.; Griffith, Elias J.; Ralph, Jason F.; Maskell, Simon

doi:10.1109/naecon.2018.8556672

Cited by 8 publications

(4 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A modified faster R-CNN was applied in the work of Terrail and Jurie [28] that showed promising performances in aerial vehicle detection. In [29], Soleimani et al proposed a text-guided detection scheme using both visual and textual features for detection. Yang et al [30] applied skip connection in their framework to merge lower and higher level features and utilized a focal loss function for vehicle detection.…”

Section: B Deep Learning Based Vehicle Detection Architecturesmentioning

confidence: 99%

Joint-SRVDNet: Joint Super Resolution and Vehicle Detection Network

et al. 2020

Self Cite

View full text Add to dashboard Cite

In many domestic and military applications, aerial vehicle detection and super-resolution algorithms are frequently developed and applied independently. However, aerial vehicle detection on superresolved images remains a challenging task due to the lack of discriminative information in the super-resolved images. To address this problem, we propose a Joint Super-Resolution and Vehicle Detection Network (Joint-SRVDNet) that tries to generate discriminative, high-resolution images of vehicles from low-resolution aerial images. First, aerial images are up-scaled by a factor of 4x using a Multi-scale Generative Adversarial Network (MsGAN), which has multiple intermediate outputs with increasing resolutions. Second, a detector is trained on super-resolved images that are upscaled by factor 4x using MsGAN architecture and finally, the detection loss is minimized jointly with the super-resolution loss to encourage the target detector to be sensitive to the subsequent super-resolution training. The network jointly learns hierarchical and discriminative features of targets and produces optimal super-resolution results. We perform both quantitative and qualitative evaluation of our proposed network on VEDAI, xView and DOTA datasets. The experimental results show that our proposed framework achieves better visual quality than the state-of-the-art methods for aerial super-resolution with 4x up-scaling factor and improves the accuracy of aerial vehicle detection.

show abstract

Section: B Deep Learning Based Vehicle Detection Architecturesmentioning

confidence: 99%

Joint-SRVDNet: Joint Super Resolution and Vehicle Detection Network

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…A modified faster R-CNN was applied in the work of Terrail et al [28] that showed promising performances in aerial vehicle detection. In [29], Soleimani et al proposed a text-guided detection scheme using both visual and textual features for detection. Yang et al [30] applied skip connection in their framework to merge lower and higher level features and utilized a focal loss function for vehicle detection.…”

Section: B Deep Learning Based Vehicle Detection Architecturesmentioning

confidence: 99%

Joint-SRVDNet: Joint Super Resolution and Vehicle Detection Network

Mostofa¹,

Ferdous²,

Riggan³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

In many domestic and military applications, aerial vehicle detection and super-resolution algorithms are frequently developed and applied independently. However, aerial vehicle detection on superresolved images remains a challenging task due to the lack of discriminative information in the superresolved images. To address this problem, we propose a Joint Super-Resolution and Vehicle Detection Network (Joint-SRVDNet) that tries to generate discriminative, high-resolution images of vehicles from low-resolution aerial images. First, aerial images are up-scaled by a factor of 4x using a Multi-scale Generative Adversarial Network (MsGAN), which has multiple intermediate outputs with increasing resolutions. Second, a detector is trained on super-resolved images that are upscaled by factor 4x using MsGAN architecture and finally, the detection loss is minimized jointly with the super-resolution loss to encourage the target detector to be sensitive to the subsequent super-resolution training. The network jointly learns hierarchical and discriminative features of targets and produces optimal super-resolution results. We perform both quantitative and qualitative evaluation of our proposed network on VEDAI, xView and DOTA datasets. The experimental results show that our proposed framework achieves better visual quality than the state-of-the-art methods for aerial super-resolution with 4x up-scaling factor and improves the accuracy of aerial vehicle detection.

show abstract

“…Object detection in satellite imagery remains an ongoing research topic. Soleimani et al [4] investigate the problem of vehicle recognition in aerial imagery using a text-guided deep convolutional neural network classifier. They used a dataset with images captured by drones or Unmanned Aerial Vehicles (UAVs).…”

Section: Related Workmentioning

confidence: 99%

Identifying Vehicles in Satellite Imagery Using Transfer Learning – Cluj-Napoca Experiment

Rădac¹

2019

Proceedings of the 18th International Conference on INFORMATICS in ECONOMY Education, Research and Business Technologies

View full text Add to dashboard Cite

This paper introduces a prototype for detecting and recognizing vehicles on satellite imagery from a city in Romania. We used a transfer learning approach, reusing layers from a convolutional neural network classifier a-priori trained with external data. For our classification process, we used the CNN structure of xView, trained with images from the VEDAI dataset. In the final data prediction step, we recognize vehicles from the city of Cluj-Napoca, using images acquired from Google Earth.

show abstract

Convolutional Neural Networks for Aerial Vehicle Detection and Recognition

Cited by 8 publications

References 18 publications

Joint-SRVDNet: Joint Super Resolution and Vehicle Detection Network

Joint-SRVDNet: Joint Super Resolution and Vehicle Detection Network

Joint-SRVDNet: Joint Super Resolution and Vehicle Detection Network

Identifying Vehicles in Satellite Imagery Using Transfer Learning – Cluj-Napoca Experiment

Contact Info

Product

Resources

About