Deep Learning for Multilabel Land Cover Scene Categorization Using Data Augmentation

Stivaktakis, Radamanthys; Tsagkatakis, Grigorios; Tsakalides, Panagiotis

doi:10.1109/lgrs.2019.2893306

Cited by 93 publications

(46 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In general, the FCN methods performed excellently in land cover classification. This is consistent with many results by researchers in the mapping field [7,[25][26][27][28][29]. Spatial features in remotely sensed data are very important in classification and are intrinsically local and spatially invariant.…”

Section: Discussionsupporting

confidence: 91%

Comparing Fully Deep Convolutional Neural Networks for Land Cover Classification with High-Spatial-Resolution Gaofen-2 Images

Han

Dian

Xia

et al. 2020

IJGI

View full text Add to dashboard Cite

Land cover is an important variable of the terrestrial ecosystem that provides information for natural resources management, urban sprawl detection, and environment research. To classify land cover with high-spatial-resolution multispectral remote sensing imagery is a difficult problem due to heterogeneous spectral values of the same object on the ground. Fully convolutional networks (FCNs) are a state-of-the-art method that has been increasingly used in image segmentation and classification. However, a systematic quantitative comparison of FCNs on high-spatial-multispectral remote imagery was not yet performed. In this paper, we adopted the three FCNs (FCN-8s, Segnet, and Unet) for Gaofen-2 (GF2) satellite imagery classification. Two scenes of GF2 with a total of 3329 polygon samples were used in the study area and a systematic quantitative comparison of FCNs was conducted with red, green, blue (RGB) and RGB+near infrared (NIR) inputs for GF2 satellite imagery. The results showed that: (1) The FCN methods perform well in land cover classification with GF2 imagery, and yet, different FCNs architectures exhibited different results in mapping accuracy. The FCN-8s model performed best among the Segnet and Unet architectures due to the multiscale feature channels in the upsampling stage. Averaged across the models, the overall accuracy (OA) and Kappa coefficient (Kappa) were 5% and 0.06 higher, respectively, in FCN-8s when compared with the other two models. (2) High-spatial-resolution remote sensing imagery with RGB+NIR bands performed better than RGB input at mapping land cover, and yet the advantage was limited; the OA and Kappa only increased an average of 0.4% and 0.01 in the RGB+NIR bands. (3) The GF2 imagery provided an encouraging result in estimating land cover based on the FCN-8s method, which can be exploited for large-scale land cover mapping in the future.

show abstract

Section: Discussionsupporting

confidence: 91%

Comparing Fully Deep Convolutional Neural Networks for Land Cover Classification with High-Spatial-Resolution Gaofen-2 Images

Han

Dian

Xia

et al. 2020

IJGI

View full text Add to dashboard Cite

show abstract

“…In fully connected layers, each node is connected with every other node from the previous layer, following the prescription of typical Multi-layer Perception architectures [20]. Fully connected and pooling layers are more frequently found in scenarios involving classification tasks where sequences of such layers are cascaded reaching the final output layer, e.g., [21,22,23]; however, since reduction of dimensionality is not required in inverse imaging problems, they are not typically employed for observation enhancement tasks.…”

Section: Deep Neural Network Paradigmsmentioning

confidence: 99%

Survey of Deep-Learning Approaches for Remote Sensing Observation Enhancement

Tsagkatakis

Aidini

Fotiadou

et al. 2019

Sensors

Self Cite

116

View full text Add to dashboard Cite

Deep Learning, and Deep Neural Networks in particular, have established themselves as the new norm in signal and data processing, achieving state-of-the-art performance in image, audio, and natural language understanding. In remote sensing, a large body of research has been devoted to the application of deep learning for typical supervised learning tasks such as classification. Less yet equally important effort has also been allocated to addressing the challenges associated with the enhancement of low-quality observations from remote sensing platforms. Addressing such channels is of paramount importance, both in itself, since high-altitude imaging, environmental conditions, and imaging systems trade-offs lead to low-quality observation, as well as to facilitate subsequent analysis, such as classification and detection. In this paper, we provide a comprehensive review of deep-learning methods for the enhancement of remote sensing observations, focusing on critical tasks including single and multi-band super-resolution, denoising, restoration, pan-sharpening, and fusion, among others. In addition to the detailed analysis and comparison of recently presented approaches, different research avenues which could be explored in the future are also discussed.

show abstract

“…For image analysis, data expansion can be an effective solution that can flip, translate or rotate existing images to create more data and make neural networks better generalized (Stivaktakis, Tsagkatakis & Tsakalides, 2019). The application of data expansion can also improve the robustness of the model.…”

Section: Data Augmentationmentioning

confidence: 99%

“…For example, the multilabel land cover scene categorization project adopts an online method for data augmentation, where each training batch is dynamically enhanced at each round of iteration. Because CNN rarelies or never processes the same example twice, dynamic expansion eliminates the memory requirements associated with more massive static datasets and enhances the generalization capabilities of the network compared to offline alternatives (Stivaktakis, Tsagkatakis & Tsakalides, 2019).…”

Section: Data Augmentationmentioning

confidence: 99%

Vehicle-Related Scene Understanding Using Deep Learning

Liu

Neuyen

Yan

2020

Communications in Computer and Information Science

View full text Add to dashboard Cite

Automated driving technology is an inevitable trend in the future development of transportation, it is also one of the eminent achievements in the matter of artificial intelligence. Primarily deep learning produces a significant contribution to the progression of automatic driving. Deep learning not only promotes autonomous vehicles to sense and identify the surrounding environment, but also identifies and classifies various information regarding to vehicles. With the upgrades and improvement of deep learning technology, it can be promptly and readily learned and employed. A large number of pretraining networks and public datasets have provided convenience for training numerous traffic scenes. Nevertheless, automated driving technology is not flexible enough to understand scenes in complex traffic environments, with regard to traffic rules and transportation facilities in various countries. There is no algorithm so far designed for all traffic scenes. In this thesis, our contributions are that we primarily deal with the issue of understanding of vehicle-related scene using deep learning. To the best of our knowledge, this is the first time that we utilize Auckland traffic environment as an analysis object for scene understanding. Moreover, automatic scene segmentation and object detection are coalesced for traffic scene understanding. The techniques based on deep learning dramatically decrease human manipulations. Furthermore, the datasets in this project provide a large amount of Auckland traffic data. Meanwhile, the performance of CNN processing is consolidated by combining with vehicle detection outcome.

show abstract

Deep Learning for Multilabel Land Cover Scene Categorization Using Data Augmentation

Cited by 93 publications

References 17 publications

Comparing Fully Deep Convolutional Neural Networks for Land Cover Classification with High-Spatial-Resolution Gaofen-2 Images

Comparing Fully Deep Convolutional Neural Networks for Land Cover Classification with High-Spatial-Resolution Gaofen-2 Images

Survey of Deep-Learning Approaches for Remote Sensing Observation Enhancement

Vehicle-Related Scene Understanding Using Deep Learning

Contact Info

Product

Resources

About