Semantic Segmentation with Extended DeepLabv3 Architecture

Yurtkulu, Salih Can; Sahin, Yusuf Huseyin; Ünal, Gözde

doi:10.1109/siu.2019.8806244

Cited by 82 publications

(54 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Mask R-CNN (He et al, 2020) is still an excellent solution of segmentation, however DeepLabV3 (Yurtkulu et al, 2019) is considered a state-ofthe-art model of human semantic segmentation.…”

Section: Related Workmentioning

confidence: 99%

“…These methods employ image segmentation to predict the alpha mattes from an RGB image input. IndexNet Matting (Yurtkulu et al, 2019) employs encoder and decoder network in order to learn index pooling and un-pooling. Context-Aware Matting (Hou & Liu, 2019) introduces double decoders to estimate alpha and foreground map.…”

Section: Related Workmentioning

confidence: 99%

“…In the first part of the proposed network (Figure 2), the Deep Learning Matting Network (DLMN-G) is introduced that takes as input an original image with a person in the foreground. Firstly, atrous convolution is applied in order to generate softsegmentation using DeepLabV3 (Yurtkulu et al, 2019). It should be noted that soft-segmentation has been used inside this network in order to automatically generate a trimap while applying erosion and dilatation.…”

Section: Deep Learning Matting Network Generatormentioning

confidence: 99%

See 2 more Smart Citations

Deep Residual Network for High-Resolution Background Matting

Abdellaoui¹,

Kachbal²

2021

STUD INFORM CONTROL

View full text Add to dashboard Cite

Image matting is one of the most important tasks in the computer vision community whose popularity has increased in recent years. This is a highly critical method in video and image editing applications, which involves the separation of the foreground from the background of an image. The previous methods provide a low accuracy when the background and foreground of an image are similar. This paper proposes an effective matting method that integrates the combination of a supervised deep learning matting network generator and a self-supervised refinement network. The generator uses a supervised encoder-decoder network for the extraction of the foreground and alpha matte from the original input image. The results obtained by this network are employed by the self-supervised refinement network to evaluate the newly created composite images and ultimately improving the matting process. The proposed method has obtained better results in comparison with other methods, which makes it more reliable.

show abstract

“…Mask R-CNN (He et al, 2020) is still an excellent solution of segmentation, however DeepLabV3 (Yurtkulu et al, 2019) is considered a state-ofthe-art model of human semantic segmentation.…”

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Deep Learning Matting Network Generatormentioning

confidence: 99%

See 1 more Smart Citation

Deep Residual Network for High-Resolution Background Matting

Abdellaoui¹,

Kachbal²

2021

STUD INFORM CONTROL

View full text Add to dashboard Cite

show abstract

“…Semantic segmentation provides a practical way to understand high spatial resolution images by assigning each pixel a semantic label, which is set building in our case. The advent of FCN (Fully convolution network) [17], Unet [18], SegNet [19], feature pyramid networks [20], PSPNet [21] and their developed network structures [22][23][24] have improved the performance in building detection significantly. Extracting the multi-scale features from high spatial resolution images is one major application area that researchers are working on, particularly in remotely sensed applications [25].…”

Section: Introductionmentioning

confidence: 99%

Res2-Unet, a New Deep Architecture for Building Detection From High Spatial Resolution Images

Chen

Wang

et al. 2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Accurate large-scale building detection is significant in monitoring urban development, map updating, change detection and digital city establishment. However, due to the complicated details of background objects in high spatial resolution remotely sensed images, the models proposed in building detection are still not performing satisfactorily. Particularly, such issue lies in the small buildings which are easily to be omitted, and the pixels in the bounding area of each building instance can be especially confusing with the background objects. Aiming to deal with such problem, we propose Res2-Unet to employ multi-scale learning at a granular level, rather than the commonly used layer-wise feature learning, to enlarge the scale of receptive fields of each bottleneck layer. It replaces the widely used 3×3 convolution on n-channel feature maps with a set of smaller groups, which are organized in a hierarchical structure to enlarge the scale-variability. The general framework is an end-to-end learning network, taking a typical semantic segmentation network structure with encoders to encode the input image into feature maps and decoders to decode the feature maps into binary segmented result image. Moreover, to enhance the building boundary generation ability of our model, a boundary loss function is proposed to improve the detection performance. The proposed framework is evaluated on three public datasets, Massachusetts building dataset, WHU East Asia Satellite dataset and WHU Aerial building dataset. It is compared with the published performances and has achieved the state-of-the-art accuracies. That verifies the robustness of the proposed framework.

show abstract

“…The dilation architecture uses dilated convolutions to preserve high-resolution feature representations. Examples of current state-of-the-art (SOTA) architectures with dilated convolutions include Deeplabv3 [15] and PSPNet [16]. The encoder-decoder architecture has downsampling and upsampling components and uses skip or lateral connections to capture wider context and recover the high-resolution feature representations.…”

Section: Introductionmentioning

confidence: 99%

Trilateral Attention Network for Real-Time Cardiac Region Segmentation

2021

View full text Add to dashboard Cite

The accurate segmentation of cardiac images into anatomically meaningful regions is critical for the extraction of quantitative cardiac indices. The common pipeline for segmentation comprises regions of interest (ROIs) localization and segmentation stages that are independent of each other and typically performed using separate models. In this paper, we propose an end-to-end network, called Trilateral Attention Network (TaNet), for real-time region localization and segmentation. TaNet has a module for ROIs localization and three segmentation pathways: spatial pathway, handcrafted pathway, and context pathway. The localization module focuses segmentation attention on the desired region while learning the context relationship between different regions in the image. The localized regions are then sent to the three pathways for segmentation. The spatial pathway, which has regular convolutional kernels, is used to extract deep features at different levels of abstraction. The handcrafted pathway, which has hand-designed convolutional kernels, is used to extract a unique set of features complementary to the deep features. Finally, the context (or global) pathway is used to enlarge the receptive field. By jointly training TaNet for localization and segmentation, TaNet achieved superior performance, in terms of accuracy and speed, when evaluated on two echocardiography datasets for cardiac region segmentation. The code and models will be made publicly available in TaNet Github page.

show abstract

Semantic Segmentation with Extended DeepLabv3 Architecture

Cited by 82 publications

References 10 publications

Deep Residual Network for High-Resolution Background Matting

Deep Residual Network for High-Resolution Background Matting

Res2-Unet, a New Deep Architecture for Building Detection From High Spatial Resolution Images

Trilateral Attention Network for Real-Time Cardiac Region Segmentation

Contact Info

Product

Resources

About