Fast Semantic Segmentation for Scene Perception

Zhang, Xuetao; Chen, Zhenxue; Wu, Q. M. Jonathan; Cai, Lei; Lü, Dan; Li, Xianming

doi:10.1109/tii.2018.2849348

Cited by 107 publications

(57 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As explained in the previous section that P-LPN is a new framework for pedestrian relative location identification in driving scene, it is neither a semantic segmentation model (e.g. PSPNet [16], SegNet [22], ICNet [24], etc.) nor a target detection model (e.g.…”

Section: A Performance Evaluation Indicatorsmentioning

confidence: 99%

“…Chen pioneered the use of Atrous convolutions and full connected Conditional Random Field (CRF) in a series of effective semantic segmentation solutions (DeeplabV1 to DeeplabV3) [17], [18]. Zhang et al proposed a more efficient asymmetric encoder-decoder structure for semantic segmentation, it has much fewer parameters than Segnet [22]. In fact, previous deep learning based semantic segmentation models suffer from low efficiency, they mainly exploit fully convolutional networks (FCNs) which is a sophisticated architecture with multiple layers of convolution, pooling, and normalization, etc.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

P-LPN: Towards Real Time Pedestrian Location Perception in Complex Driving Scenes

Zhao

et al. 2020

IEEE Access

View full text Add to dashboard Cite

Semantic segmentation is one of the most critical modules in road scene understanding. In this paper, we focus on the challenging task of pedestrian's relative location perception in the semantic graph of complex driving scenes. Prevalent research on semantic segmentation mainly concentrate on improving the segmentation accuracy with less attention paid to computational efficiency. Furthermore, little effort has been made in pedestrian location perception in complex driving scenes. For example, current semantic segmentation methods classify all pedestrians as a mono category, regardless of whether the pedestrians are penetrating into the vehicular lane or standing still in the safe sidewalk area. We propose a pedestrian location perception network (P-LPN). P-LPN can produce real-time semantic segmentation while simultaneously providing location inference for each pedestrian in semantic maps. This enables autonomous driving system to categorize pedestrians into different safety levels. We comprehensively evaluated P-LPN on CityScapes benchmark through comparative studies. Our proposal achieved competitive performance in both accuracy and efficiency. It yields quality inference with real-time speed at ∼22 fps.

show abstract

Section: A Performance Evaluation Indicatorsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

P-LPN: Towards Real Time Pedestrian Location Perception in Complex Driving Scenes

Zhao

et al. 2020

IEEE Access

View full text Add to dashboard Cite

show abstract

“…However, fisheye images are not able to segment images, especially on edges. New and more improved methods [3,[46][47][48][49] are offering more enhanced segmentation, though our research's key objective is to reduce the size of a segmentation network without losing feature details. Recently, a very promising CNN architecture was released named DeepLabV3 [4].…”

Section: Related Workmentioning

confidence: 99%

ECRU: An Encoder-Decoder Based Convolution Neural Network (CNN) for Road-Scene Understanding

Yasrab

2018

J. Imaging

View full text Add to dashboard Cite

This research presents the idea of a novel fully-Convolutional Neural Network (CNN)-based model for probabilistic pixel-wise segmentation, titled Encoder-decoder-based CNN for Road-Scene Understanding (ECRU). Lately, scene understanding has become an evolving research area, and semantic segmentation is the most recent method for visual recognition. Among vision-based smart systems, the driving assistance system turns out to be a much preferred research topic. The proposed model is an encoder-decoder that performs pixel-wise class predictions. The encoder network is composed of a VGG-19 layer model, while the decoder network uses 16 upsampling and deconvolution units. The encoder of the network has a very flexible architecture that can be altered and trained for any size and resolution of images. The decoder network upsamples and maps the low-resolution encoder’s features. Consequently, there is a substantial reduction in the trainable parameters, as the network recycles the encoder’s pooling indices for pixel-wise classification and segmentation. The proposed model is intended to offer a simplified CNN model with less overhead and higher performance. The network is trained and tested on the famous road scenes dataset CamVid and offers outstanding outcomes in comparison to similar early approaches like FCN and VGG16 in terms of performance vs. trainable parameters.

show abstract

“…ESPNet [16] decomposes efficiently a standard convolutional kernel into a point-wise convolution and multiple dilated convolutions with a spatial pyramids structure, achieving large receptive fields and a good trade-off between speed and accuracy. Zhang et al construct a fast semantic segmentation network named FSSNet [17] by multiple effective blocks to achieve high accuracy performance with only 0.2M parameters, but it did not have a good accuracy performance for similar inter-class objects. ShuffleNet [18], a light-weight image classification network, uses point-wise group convolution to reduce feature map dimensions first, and increase the dimensions later.…”

Section: Introductionmentioning

confidence: 99%

“…To further reduce parameters and improve accuracy performance, we propose the efficient fast semantic segmentation network named EFSNet, which adopts channel shuffle dilated convolution (CSDC) to perform fast and accurate segmentation. Unlike CDB [17], our proposed CSDC not only enlarges the receptive field but also reduces the number of FLOPs. It uses group convolution [18] to first reduce and then increase the number of feature map channels, uses dilated convolution [19] to enlarge the receptive field, and uses channel shuffle [18] to alleviate the loss of accuracy.…”

Section: Introductionmentioning

confidence: 99%

Efficient Fast Semantic Segmentation Using Continuous Shuffle Dilated Convolutions

Wang

2020

IEEE Access

View full text Add to dashboard Cite

It is difficult for many semantic segmentation methods to perform useful inferences under extremely resource-constrained devices; therefore, an efficient fast semantic segmentation network (EFSNet) that employs a continuous shuffle dilated convolution (CSDC) and a up-sampling module is proposed in this paper. First, the number of parameters is reduced by group convolutions and the receptive field is enlarged by dilated convolution. Second, a up-sampling module is proposed for reducing noise and increasing inference speed. Finally, with the above CSDC module and up-sampling module, the EFSNet based on a fast semantic segmentation network is obtained. Our experiments on the CamVid and Cityscapes datasets show that the mean intersection over union (mIoU) values obtained by the proposed EFSNet with only 173 k parameters are 61.1 % and 61.9 %, respectively. The method is able to run in real-time at a speed of 332 frames per second (FPS) and 107 FPS on a single NVIDIA Titan Xp GPU. Compared with several existing methods, our network achieves efficient segmentation with low resource consumption.

show abstract

Fast Semantic Segmentation for Scene Perception

Cited by 107 publications

References 26 publications

P-LPN: Towards Real Time Pedestrian Location Perception in Complex Driving Scenes

P-LPN: Towards Real Time Pedestrian Location Perception in Complex Driving Scenes

ECRU: An Encoder-Decoder Based Convolution Neural Network (CNN) for Road-Scene Understanding

Efficient Fast Semantic Segmentation Using Continuous Shuffle Dilated Convolutions

Contact Info

Product

Resources

About