We propose a learning algorithm to design a lightweight neural multiplexer that given the input and computational resource requirements, calls the model that will consume the minimum compute resources for a successful...
详细信息
ISBN:
(纸本)9781728197104
We propose a learning algorithm to design a lightweight neural multiplexer that given the input and computational resource requirements, calls the model that will consume the minimum compute resources for a successful inference. Mobile devices can use the proposed algorithm to offload the hard inputs to the cloud while inferring the easy ones locally. Besides, in the large scale cloud-based intelligent applications, instead of replicating the most-accurate model, a range of small and large models can be multiplexed from depending on the input's complexity which will save the cloud's computational resources. The input complexity or hardness is determined by the number of models that can predict the correct label. For example, if no model can predict the label correctly, then the input is considered as the hardest. The proposed algorithm allows the mobile device to detect the inputs that can be processed locally and the ones that require a larger model and should be sent a cloud server. Therefore, the mobile user benefits from not only the local processing but also from an accurate model hosted on a cloud server. Our experimental results show that the proposed algorithm improves mobile's model accuracy by 8.52% which is because of those inputs that are properly selected and offloaded to the cloud server. In addition, it saves the cloud providers' compute resources by a factor of 2.85x as small models are chosen for easier inputs.
Classic high-accuracy semantic segmentation models typically come with a large number of parameters, making them unsuitable for deployment on driverless platforms with limited computational power. To strike a balance ...
详细信息
Classic high-accuracy semantic segmentation models typically come with a large number of parameters, making them unsuitable for deployment on driverless platforms with limited computational power. To strike a balance between accuracy and limited computational budget, and enable the use of the classic segmentation model UNet in unmanned driving scenarios, this paper proposes a multi-unit stacked architecture (MSA), namely, MSA-Net, based on UNet and ShuffleNetv2. First, MSA-Net replaces the convolution blocks in the UNet encoder and decoder with stacked basic ShuffleNetv2 units, which greatly reduces computational cost while maintaining high segmentation accuracy. Second, MSA-Net designs enhanced skip connections using pointwise convolution and convolutional block attention (CBAM) to aid the decoder in selecting more relevant and valuable information. Third, MSA-Net proposes multi-scale internal connections to extend the receptive fields of encoder and decoder with little increase in model parameters. The comprehensive experiments show MSANet achieves an optimal balance on the Cityscapes dataset between accuracy and model complexity, with strong generalization on the enhanced PASCAL VOC 2012 dataset. MSA-Net achieves a mean intersection over union (mIoU) of 73.6% and an inference speed of 31.0 frames per second (FPS) on the Cityscapes test dataset. We also propose two other MSA-Net models of different sizes, providing more options for resource-constrained inference.
暂无评论