DSpace-CRIS

Now showing 1 - 3 of 3

Some of the metrics are blocked by your
consent settings
Transformer-based decoder designs for semantic segmentation on remotely sensed images
(2021-12-01)
Panboonyuen, Teerapong
;
Jitkajornwanich, Kulsawasd
;
Lawawirojwong, Siam
;
Srestasathiern, Panu
;
Vateekul, Peerapon
Transformers have demonstrated remarkable accomplishments in several natural language processing (NLP) tasks as well as image processing tasks. Herein, we present a deep-learning (DL) model that is capable of improving the semantic segmentation network in two ways. First, utilizing the pre-training Swin Transformer (SwinTF) under Vision Transformer (ViT) as a backbone, the model weights downstream tasks by joining task layers upon the pretrained encoder. Secondly, decoder designs are applied to our DL network with three decoder designs, U-Net, pyramid scene parsing (PSP) network, and feature pyramid network (FPN), to perform pixel-level segmentation. The results are compared with other image labeling state of the art (SOTA) methods, such as global convolutional network (GCN) and ViT. Extensive experiments show that our Swin Transformer (SwinTF) with decoder designs reached a new state of the art on the Thailand Isan Landsat-8 corpus (89.8% F1 score), Thailand North Landsat-8 corpus (63.12% F1 score), and competitive results on ISPRS Vaihingen. Moreover, both our best-proposed methods (SwinTF-PSP and SwinTF-FPN) even outperformed SwinTF with supervised pre-training ViT on the ImageNet-1K in the Thailand, Landsat-8, and ISPRS Vaihingen corpora.
Some of the metrics are blocked by your
consent settings
Semantic labeling in remote sensing corpora using feature fusion-based enhanced global convolutional network with high-resolution representations and depthwise atrous convolution
(2020-04-01)
Panboonyuen, Teerapong
;
Jitkajornwanich, Kulsawasd
;
Lawawirojwong, Siam
;
Srestasathiern, Panu
;
Vateekul, Peerapon
One of the fundamental tasks in remote sensing is the semantic segmentation on the aerial and satellite images. It plays a vital role in applications, such as agriculture planning, map updates, route optimization, and navigation. The state-of-the-art model is the Enhanced Global Convolutional Network (GCN152-TL-A) from our previous work. It composes two main components: (i) the backbone network to extract features and (ii) the segmentation network to annotate labels. However, the accuracy can be further improved, since the deep learning network is not designed for recovering low-level features (e.g., river, low vegetation). In this paper, we aim to improve the semantic segmentation network in three aspects, designed explicitly for the remotely sensed domain. First, we propose to employ a modern backbone network called "High-Resolution Representation (HR)" to extract features with higher quality. It repeatedly fuses the representations generated by the high-to-low subnetworks with the restoration of the low-resolution representations to the same depth and level. Second, "Feature Fusion (FF)" is added to our network to capture low-level features (e.g., lines, dots, or gradient orientation). It fuses between the features from the backbone and the segmentation models, which helps to prevent the loss of these low-level features. Finally, "Depthwise Atrous Convolution (DA)" is introduced to refine the extracted features by using four multi-resolution layers in collaboration with a dilated convolution strategy. The experiment was conducted on three data sets: two private corpora from Landsat-8 satellite and one public benchmark from the "ISPRS Vaihingen" challenge. There are two baseline models: the Deep Encoder-Decoder Network (DCED) and our previous model. The results show that the proposed model significantly outperforms all baselines. It is the winner in all data sets and exceeds more than 90% of F1: 0.9114, 0.9362, and 0.9111 in two Landsat-8 and ISPRS Vaihingen data sets, respectively. Furthermore, it achieves an accuracy beyond 90% on almost all classes.
Some of the metrics are blocked by your
consent settings
Combining attentional cnn and gru networks for ocean current prediction based on hf radar observations
(2019-10-23)
Thongniran, Nathachai
;
Jitkajornwanich, Kulsawasd
;
Lawawirojwong, Siam
;
Srestasathiern, Panu
;
Vateekul, Peerapon
Lately, CNN-GRU demonstrates the ability of deep learning techniques on ocean surface current prediction. Improvement of the current prediction model creates positive impact on variety of marine activities, such as search-and-rescue, disaster monitoring and power forecasting. Deep learning techniques was successfully deployed to improve model performance in many areas due to their ability to handle enormous amounts of information in a variety of inputs and their huge growth in recent years. Latest ocean current prediction employed a combination of two mature techniques, which are Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU), to capture spatial and temporal characteristics of its nature. However, there is still room for improvement due to many modern techniques that still have not been employed, and domain knowledge in oceanic, such as lunar illumination, is not taken into account to improve prediction performance. This paper introduces the ocean surface prediction model that employs soft attention mechanism, transfer learning, and incorporation of domain knowledge inputs which are month number, lunar effect, and hour number. An experimental dataset from 2014 to 2016, provided by GISTDA, is collected by using high frequency (HF) radar stations located along the coastal Gulf of Thailand. The experiment compares an existing CNN-GRU and our proposed model. The result shows an improvement of the prediction model in terms of RMSE by 2.57%, and 3.44% on U and V components.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results