An Adaptive Multiscale Fusion Network Based on Regional Attention for Remote Sensing Images

Lu, Wanzhen; Liang, Longxue; Wu, Xiaosuo; Wang, Xiaoyu; Cai, Jiali

doi:10.1109/access.2020.3000425

Cited by 4 publications

(4 citation statements)

References 27 publications

(56 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to meet the real-time distributed storage network education big data, and be good at log analysis, it can further analyze the user behavior data generated during the learning process of students and teachers [34][35][36]. Hadoop is a distributed system infrastructure developed by the Apache Foundation.…”

Section: ) Hadoop Clustermentioning

confidence: 99%

Regional Network Education Information Collection Platform for Smart Classrooms based on Big Data Technology

Sharma

2021

IJCAI

View full text Add to dashboard Cite

Big data technology plays an important role in optimizing the education intelligence by enhancing the learning experience through novel assessment strategies and predictive teaching. This work aims at improving the learning efficiency of users and increase the effective learning quality of the smart classroom concept using the big data technology. A regional network education information collection platform based on big data technology has been developed, which can collect student learning data for subsequent analysis and processing. The software architecture of this platform is mainly divided into basic layer, platform layer and access layer. The physical structure mainly includes web servers and Hadoop clusters. The big data acquisition platform of this education has excellent acquisition performance, and the data acquisition of different data item field length is larger than the expected index. The outcomes obtained reveals that the data item field is less than 20 and the acquisition amount is 145% of the expected result. When the data item field is between 20 and 40, the acquisition amount is 137% of the expected result. The collection is 116% of the expected result, when the data item field is between 40 and 50. When the data item field is between 50 and 60, the collection is 103% of the expected result. The comparison of the regular teaching method with the smart classroom based concept revealed that the big data platform is 87% better than the normal regular teaching methods in terms of study material as well as meaningful teacher-student interaction. The research on education big data can provide better education services for educational activities, drive the reform of teaching mode and optimize the teaching methods of education enabling the smart classroom concept.Povzetek: Predstavljena je regionalna platforma za upravljanje pametne učilnice na osnovi tehnologije velikih podatkov.

show abstract

Section: ) Hadoop Clustermentioning

confidence: 99%

Regional Network Education Information Collection Platform for Smart Classrooms based on Big Data Technology

Sharma

2021

IJCAI

View full text Add to dashboard Cite

show abstract

“…The GCN-based semantic segmentation model achieves better results in image segmentation tasks [34,57,58]. We improve the original GCN, as shown in Figure 3, from a single-size convolutional kernel to a two-by-two combination of three convolutional kernels with different sizes, 3 × 3, 5 × 5, and 7 × 7.…”

Section: Hybrid-scale Gcnmentioning

confidence: 99%

HA-Net: A Lake Water Body Extraction Network Based on Hybrid-Scale Attention and Transfer Learning

2021

View full text Add to dashboard Cite

Due to the large quantity of noise and complex spatial background of the remote sensing images, how to improve the accuracy of semantic segmentation has become a hot topic. Lake water body extraction is crucial for disaster detection, resource utilization, and carbon cycle, etc. The the area of lakes on the Tibetan Plateau has been constantly changing due to the movement of the Earth’s crust. Most of the convolutional neural networks used for remote sensing images are based on single-layer features for pixel classification while ignoring the correlation of such features in different layers. In this paper, the two-branch encoder is presented, which is a multiscale structure that combines the features of ResNet-34 with a feature pyramid network. Secondly, adaptive weights are distributed to global information using the hybrid-scale attention block. Finally, PixelShuffle is used to recover the feature maps’ resolution, and the densely connected block is used to refine the boundary of the lake water body. Likewise, we transfer the best weights which are saved on the Google dataset to the Landsat-8 dataset to ensure that our proposed method is robust. We validate the superiority of Hybrid-scale Attention Network (HA-Net) on two given datasets, which were created by us using Google and Landsat-8 remote sensing images. (1) On the Google dataset, HA-Net achieves the best performance of all five evaluation metrics with a Mean Intersection over Union (MIoU) of 97.38%, which improves by 1.04% compared with DeepLab V3+, and reduces the training time by about 100 s per epoch. Moreover, the overall accuracy (OA), Recall, True Water Rate (TWR), and False Water Rate (FWR) of HA-Net are 98.88%, 98.03%, 98.24%, and 1.76% respectively. (2) On the Landsat-8 dataset, HA-Net achieves the best overall accuracy and the True Water Rate (TWR) improvement of 2.93% compared to Pre_PSPNet, which proves to be more robust than other advanced models.

show abstract

“…Here, * denotes the convolution operation. Concat is an increase in the number of channels [33], while the information under each feature is not increased. We set the number of channels of output feature maps to 256 in our experiments, so all convolutional layers have 256 channel outputs.…”

Section: A Feature Pyramid Networkmentioning

confidence: 99%

A Lightweight Feature Fusion Single Shot Multibox Detector for Garbage Detection

Wang

2020

IEEE Access

View full text Add to dashboard Cite

To solve the problems of a poor manual garbage sorting environment, including heavy tasks and low sorting efficiency, we propose the Lightweight Feature Fusion Single Shot Multibox Detector (L-SSD) algorithm to realize intelligent trash classification and recognition. Since waste has a small volume and the image resolution of garbage is always low, the algorithm that we propose is an enhanced single shot multibox detector (SSD) with a lightweight and novel feature fusion module. This SSD can significantly improve the performance of rubbish detection. In this feature fusion module, features from different layers with different scales are connected in series. A new feature pyramid was generated by using downsampling blocks, which will be fed to appointed multibox detectors to predict the final detection results. Due to the extremely unbalanced ratio of positive samples to negative samples, which leads to a low accuracy of SSD, Focal Loss using balanced cross-entropy is employed, which is provided by easy examples that corresponds to difficult samples with a decline in the loss weight. Thus, the training is biased towards meaningful samples. We have replaced the backbone network of VGG16 with ResNet-101 to achieve more accurate detection. We analyzed the performance of a nonmaximum suppression (NMS) algorithm and discovered that Soft-NMS was more suitable for learning better image representations. The strategy of Soft-NMS is to suppress the undesirable detection box rather than remove it completely. The experimental results show that the L-SSD exceeds a large number of state-of-the-art object detection algorithms in both accuracy and speed.

show abstract

An Adaptive Multiscale Fusion Network Based on Regional Attention for Remote Sensing Images

Cited by 4 publications

References 27 publications

Regional Network Education Information Collection Platform for Smart Classrooms based on Big Data Technology

Regional Network Education Information Collection Platform for Smart Classrooms based on Big Data Technology

HA-Net: A Lake Water Body Extraction Network Based on Hybrid-Scale Attention and Transfer Learning

A Lightweight Feature Fusion Single Shot Multibox Detector for Garbage Detection

Contact Info

Product

Resources

About