Efficient Fast Semantic Segmentation Using Continuous Shuffle Dilated Convolutions

Hu, Xuegang; Wang, Haibo

doi:10.1109/access.2020.2987080

Cited by 21 publications

(3 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This paper uses DUTS [7] , ECSSD [8] , HKU-IS [9] and PASCAL-S [10] as experimental datasets. The introduction of each data set is as follows: DUTS contains 10553 training maps and 5019 testing maps.…”

Section: Datasetmentioning

confidence: 99%

Hierarchical feature learning network for salient object detection

Pan,

Pan

2023

Second International Conference on Applied Statistics, Computational Mathematics, and Software Engineering (ASCMSE 2023)

View full text Add to dashboard Cite

How to effectively use multi-level features is the key to improve the performance of salient object detection (SOD) model. Many existing methods use the same or similar learning strategies to deal with multi-level features, while there are relatively few studies on using different learning strategies to deal with multi-level features. In addition, how to fully fuse multi-level features is an important process to obtain accurate saliency map, A novel hierarchical feature learning network (HFLNet) is proposed to realize salient object detection. The whole detection process can be divided into three stages. Firstly, ResNet-50 is used as the backbone network to extract multi-level features; Then, multi-level features are processed by residual denoising module, interaction module and adjacent pyramid module respectively; After obtaining the further learned multi-level features, a cross fusion module is proposed to progressively fuse the multi-level features from top to bottom and bottom to top, and finally fuse the results of the above two paths to obtain the final prediction results. During the training of network model, a hybrid loss function is applied to assist the training. The proposed method is compared with nine related methods on four public datasets. The experimental results show that the method not only improves the four important quantitative performance metrics, especially the MAE, but also can predict more complete saliency maps.

show abstract

Section: Datasetmentioning

confidence: 99%

Hierarchical feature learning network for salient object detection

Pan,

Pan

2023

Second International Conference on Applied Statistics, Computational Mathematics, and Software Engineering (ASCMSE 2023)

View full text Add to dashboard Cite

show abstract

“…ESCNet [ 19 ] utilizes an efficient spatio-channel dilated convolution (ESC) module, which is an efficient multi-level dilated convolution module, to accomplish various receptive fields with reduced network parameters and computational complexity. EFSNet [ 42 ] propose the continuous shuffle dilated convolution (CSDC) module for less calculational effort. Efficient semantic segmentation methods focus on speed or low usage of memories.…”

Section: Related Workmentioning

confidence: 99%

Knowledge Distillation for Semantic Segmentation Using Channel and Spatial Correlations and Adaptive Cross Entropy

Park

Heo

2020

Sensors

View full text Add to dashboard Cite

In this paper, we propose an efficient knowledge distillation method to train light networks using heavy networks for semantic segmentation. Most semantic segmentation networks that exhibit good accuracy are based on computationally expensive networks. These networks are not suitable for mobile applications using vision sensors, because computational resources are limited in these environments. In this view, knowledge distillation, which transfers knowledge from heavy networks acting as teachers to light networks as students, is suitable methodology. Although previous knowledge distillation approaches have been proven to improve the performance of student networks, most methods have some limitations. First, they tend to use only the spatial correlation of feature maps and ignore the relational information of their channels. Second, they can transfer false knowledge when the results of the teacher networks are not perfect. To address these two problems, we propose two loss functions: a channel and spatial correlation (CSC) loss function and an adaptive cross entropy (ACE) loss function. The former computes the full relationship of both the channel and spatial information in the feature map, and the latter adaptively exploits one-hot encodings using the ground truth labels and the probability maps predicted by the teacher network. To evaluate our method, we conduct experiments on scene parsing datasets: Cityscapes and Camvid. Our method presents significantly better performance than previous methods.

show abstract

“…The deep convolution neural network (DCNN) achieves performance that traditional algorithms cannot match on many tasks, such as speech recognition [1], object detection [2], semantic segmentation [3], and so forth. However, these networks are computationally expensive (over billions of multiply-and-accumulate) and have a large number of parameters (more than 10 million parameters), making them difficult to implement on embedded platforms.…”

Section: Introductionmentioning

confidence: 99%

A digital signal processor‐efficient accelerator for depthwise separable convolution

Huang

Liu

et al. 2022

Electronics Letters

View full text Add to dashboard Cite

Recent researches on deep convolution neural networks have proposed some compact networks, such as MobileNet, but its main computation, depthwise separable convolution (DWC), which reduces the reusable data and improves the requirement of data loading efficiency. Although DWC can effectively reduce the amount of network computation, it needs a special accelerator to enhance the inference speed. This paper proposes a high-performance accelerator for DWC based on the commonly used acceleration platform field-programmable gate array. The proposed accelerator supports the computation of both standard convolutions (SCs) and DWC as well as two activation functions. In addition, two data storage formats are used to maintain the data loading efficiency for different input requirements of SC and DWC under high parallelism. Furthermore, a processing unit that can execute two 8 × 8bit multiplications inside one digital signal processor (DSP) is designed to make the best use of the DSP hardware resources. Finally, the accelerator is implemented on ZYNQ ZC706 at 200 MHz. Consuming only 392 DSPs, the accelerator achieves 134.5 giga operations per second (GOPS) and 209.4 frames per second (FPS) on MobileNet V1 as well as 96.4 GOPS and 250.4 FPS on MobileNet V2. Experimental results demonstrate that this design provides a better DSP efficiency than previous works.

show abstract

Efficient Fast Semantic Segmentation Using Continuous Shuffle Dilated Convolutions

Cited by 21 publications

References 27 publications

Hierarchical feature learning network for salient object detection

Hierarchical feature learning network for salient object detection

Knowledge Distillation for Semantic Segmentation Using Channel and Spatial Correlations and Adaptive Cross Entropy

A digital signal processor‐efficient accelerator for depthwise separable convolution

Contact Info

Product

Resources

About