LinkNet: Exploiting encoder representations for efficient semantic segmentation

By Abhishek Chaurasia, Eugenio Culurciello. Visual Communications and Image Processing, 2017

1,712 citations recorded by the source corpus.

Pixel-wise semantic segmentation for visual scene understanding not only needs to be accurate, but also efficient in order to find any use in real-time application. Existing algorithms even though are accurate but they do not focus on utilizing the parameters of neural network efficiently. As a result they are huge in terms of parameters and number of operations; hence slow too. In this paper, we propose a novel deep neural network architecture which allows it to learn without any significant increase in number of parameters. Our network uses only 11.5 million parameters and 21.2 GFLOPs for processing an image of resolution 3 × 640 × 360. It gives state-of-the-art performance on CamVid and comparable results on Cityscapes dataset. We also compare our networks processing time on NVIDIA GPU and embedded system device with existing state-of-the-art architectures for different image resolutions.

View Paper: https://www.semanticscholar.org/paper/2984e527510e7e7c0eadb339a6172c05a6919b5e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

UNet++: Redesigning Skip Connections to Exploit Multiscale Features in Image Segmentation (2019, methodology): “The skip connections have shown to help recover the full spatial resolution, making fully convolutional methods suitable for semantic segmentation [40]–[43].”
3D Anisotropic Hybrid Network: Transferring Convolutional Features from 2D Images to 3D Anisotropic Volumes (2017, methodology): “Feature summation is used instead of concatenation as in [3] to support more feature maps with less memory consumption.”
Angiodysplasia Detection and Localization Using Deep Convolutional Neural Networks (2018, methodology): “2.2 Model architecture and training In this work we evaluate 4 different deep architectures for segmentation: U-Net (Ronneberger et al., 2015; Iglovikov et al., 2017a), 2 modifications of TernausNet (Iglovikov and Shvets, 2018), and AlbuNet34, a modifications of LinkedNet (Chaurasia and Culurciello,”
Using labeled data to evaluate change detectors in a multivariate streaming environment (2009, unknown): “And FastICENet is close to CGNet [31] , LinkNet [32] and PIDNet-S [20] in accuracy, but our segmentation speed is far ahead of these networks.”
In Defense of Pre-Trained ImageNet Architectures for Real-Time Semantic Segmentation of Road-Driving Images (2019, methodology): “LinkNet [2] uses a small ResNet-18 backbone and a lightweight decoder to achieve satisfying performance-speed ratio.”
D-LinkNet: LinkNet with Pretrained Encoder and Dilated Convolution for High Resolution Satellite Imagery Road Extraction (2018, methodology): “Most latest excellent semantic segmentation networks [13, 14, 15, 16] are improved versions of FCN.”
Automatic Instrument Segmentation in Robot-Assisted Surgery Using Deep Learning (2018, methodology): “In this work we evaluate 4 different deep architectures for segmentation: UNet [9, 18], 2 modifications of TernausNet [11], and a modification of LinkNet [4].”
Weakly-Supervised Continual Learning for Class-Incremental Segmentation (2022, methodology): “For all experiments, we use a LinkNet [15] architecture trained using Adam optimizer with a learning rate of 10−4 during 10 pseudo-epochs.”
Image Segmentation Using Deep Learning: A Survey (2020, background): “Several other works adopt transposed convolutions, or encoder-decoders for image segmentation, such as Stacked Deconvolutional Network (SDN) [43], Linknet [44], W-”
Satlas: A Large-Scale, Multi-Task Dataset for Remote Sensing Image Understanding (2022, methodology): “and polylines, we compare PSPNet [55], LinkNet [14], and DeepLabv3 [15].”
LoveDA: A Remote Sensing Land-Cover Dataset for Domain Adaptive Semantic Segmentation (2021, unknown): “Specifically, the selected networks were: UNet[32], UNet++[54], LinkNet[3], DeepLabV3+[5], PSPNet[52], FCN8S[19], PAN[17], Semantic-FPN[16], HRNet[45], and FarSeg[53].”