Convolutional Neural Network with Dilated Anchors for Object Detection in Very High Resolution Satellite Images

Laban, Noureldin; Abdellatif, Bassam; Ebeid, Hala M.; Shedeed, Howida A.; Tolba, Mohamed F.

doi:10.1109/icces48960.2019.9068145

Cited by 8 publications

(4 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Object detection may be broken down into its component parts, which include categorising things and pinpointing their locations in pictures. Thus far, research efforts have been split towards optimising either one of these activities independently or both of them jointly [1] [2] If satellite photos are our data, then the metadata that characterise them are our "data for the data. "When did you take these pictures?…”

Section: Introductionmentioning

confidence: 99%

Optimization of YOLOV7's hyper parameters for simultaneous object detection in satellite imagery

Singh

2023

Preprint

View full text Add to dashboard Cite

Object detection is crucial for computer vision applications that use satellite imagery, such as precision agriculture, urban planning, and military applications.Recognizing objects in satellite images is challenging due to numerous factors, including the sheer number of objects, the variety in their positions, the range in their sizes, the quality of the lighting, and the existence of a dense background.Background complexity, differences in data capture geometry, geography, and illumination, and an abundance of different types of objects all contribute to making automatic detection in satellite images particularly difficult.There have been many advancements in object detection methods over time, including YOLO and its variations, CNN and its offshoots, DETR and its offshoots, and so on; nonetheless, it is still required to test these methods on the requisite data set to determine their true efficacy.Researchers have investigated the idea of autonomously detecting structures, automobiles, and other things to reduce the risk of human error and speed up the procedure.Improvements in deep learning algorithms and hardware systems have allowed us to accurately identify a broader range of objects in ultra-high-resolution satellite imagery.Through parameter adjustment and analysis of results on the Xview dataset, we determine the most effective technique for multiple item detection and compare it to other models.

show abstract

Section: Introductionmentioning

confidence: 99%

Optimization of YOLOV7's hyper parameters for simultaneous object detection in satellite imagery

Singh

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…The concepts of Anchors are introduced in this model. The main function of the sliding window is Anchor (Laban et al 2019). As a result, this algorithm is resistant to translations, and translational invariance is one of its important qualities (Soni 2019 Kamath et al 2020).…”

Section: Introductionmentioning

confidence: 99%

A Real Time Object Detection in Integral Part of Computer Vision using Novel Image Classification of Faster R-CNN Algorithm over Fast R-CNN Algorithm

2022

pnr

View full text Add to dashboard Cite

The objective of the work is to increase the precision of object detection using novel image classification using machine learning algorithms. Materials and Methods: The categorising is performed by adopting a sample size of n = 10 in Faster R-CNN (RPN) and sample size n = 10 in Fast R-CNN (Selective Search) algorithms with a sample size = 10 and the G-Power analysis was carried out with 80% and confidence interval 95%. Results and Discussion: The observation of the outcomes shows that the Faster R-CNN using region proposal networks has a high accuracy of 81.72% in comparison with the Selective Search based Fast R-CNN 79.61%. A statistically significant difference exists between the research groups with p=0.028 (2 tailed) (p<0.05). Conclusion: Detection of objects with high accuracy using machine learning algorithms shows that the regional proposal network based Faster R-CNN generates higher accuracy than the Selective search (Fast RCNN) algorithm.

show abstract

“…6,7 These algorithms include Region-CNN (R-CNN), Spatial Convolutional Hierarchy Networks (SPP-Net), Fast R-CNN, Faster R-CNN, You Only Look Once (YOLO), and Multi-Shot MultiBox Detector (SSD). 6,8 However, compared to natural images, CNN-based detection methods have several limitations. With remote sensing images, targets must be detected from multiple scenes, which makes detecting objects more difficult.…”

Section: Introductionmentioning

confidence: 99%

“…A CNN in deep learning methods has been used extensively in object detection 6,7 . These algorithms include Region‐CNN (R‐CNN), Spatial Convolutional Hierarchy Networks (SPP‐Net), Fast R‐CNN, Faster R‐CNN, You Only Look Once (YOLO), and Multi‐Shot MultiBox Detector (SSD) 6,8 . However, compared to natural images, CNN‐based detection methods have several limitations.…”

Section: Introductionmentioning

confidence: 99%

Toward object alphabet augmentation for object detection in very high‐resolution satellite images

Laban

Abdellatif

Ebeid

et al. 2021

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

Object detection in very high-resolution satellite images has become an important tool in many fields. So there is a vital need to build a more precise and accurate detector. There is a great challenge to detect very small and condensed objects with their different semantics shapes. Recently, convolutional neural networks with different structures have achieved awesome performance. We propose object alphabet augmentation method that extracts object instants during the training process to form an alphabet for these objects in a newly generated dataset. The proposed method uses the DarkNet-53 framework to get the main features for each object using two new generated datasets for alphabet augmentation up-down where objects are placed in up down, and alphabet augmentation horizontal where objects are placed in their original setting. Experiments are conducted on datasets randomly generated from the DOTA dataset. The experimental results show that the proposed method has improved the accuracy of detection of the target objects of the DOTA dataset for most object classes, especially for the up-down dataset.

show abstract

Convolutional Neural Network with Dilated Anchors for Object Detection in Very High Resolution Satellite Images

Cited by 8 publications

References 25 publications

Optimization of YOLOV7's hyper parameters for simultaneous object detection in satellite imagery

Optimization of YOLOV7's hyper parameters for simultaneous object detection in satellite imagery

A Real Time Object Detection in Integral Part of Computer Vision using Novel Image Classification of Faster R-CNN Algorithm over Fast R-CNN Algorithm

Toward object alphabet augmentation for object detection in very high‐resolution satellite images

Contact Info

Product

Resources

About