Probability-Based Channel Pruning for Depthwise Separable Convolutional Networks

Channel pruning can reduce memory consumption and running time with least performance damage,and is one of the most important techniques in network compression.However,existing channel pruning methods mainly focus on the pruning of standard convolutional networks,and they rely intensively on time-consuming fine-tuning to achieve the performance improvement.To this end,we present a novel efficient probability-based channel pruning method for depthwise separable convolutional networks.Our method leverages a new simple yet effective probability-based channel pruning criterion by taking the scaling and shifting factors of batch normalization layers into consideration.A novel shifting factor fusion technique is further developed to improve the performance of the pruned networks without requiring extra time-consuming fine-tuning.We apply the proposed method to five representative deep learning networks,namely MobileNetV1,MobileNetV2,ShuffleNetV1,ShuffleNetV2,and GhostNet,to demonstrate the efficiency of our pruning method.Extensive experimental results and comparisons on publicly available CIFAR10,CIFAR100,and ImageNet datasets validate the feasibility of the proposed method.

Pedestrian and Vehicle Detection Based on Pruning YOLOv4 with Cloud-Edge Collaboration

Nowadays,the rapid development of edge computing has driven an increasing number of deep learning applications deployed at the edge of the network,such as pedestrian and vehicle detection,to provide efficient intelligent services to mobile users.However,as the accuracy requirements continue to increase,the components of deep learning models for pedestrian and vehicle detection,such as YOLOv4,become more sophisticated and the computing resources required for model training are increasing dramatically,which in turn leads to significant challenges in achieving effective deployment on resource-constrained edge devices while ensuring the high accuracy performance.For addressing this challenge,a cloud-edge collaboration-based pedestrian and vehicle detection framework is proposed in this paper,which enables sufficient training of models by utilizing the abundant computing resources in the cloud,and then deploying the well-trained models on edge devices,thus reducing the computing resource requirements for model training on edge devices.Furthermore,to reduce the size of the model deployed on edge devices,an automatic pruning method combines the convolution layer and BN layer is proposed to compress the pedestrian and vehicle detection model size.Experimental results show that the framework proposed in this paper is able to deploy the pruned model on a real edge device,Jetson TX2,with 6.72 times higher FPS.Meanwhile,the channel pruning reduces the volume and the number of parameters to 96.77%for the model,and the computing amount is reduced to 81.37%.

A Real-Time Small Target Vehicle Detection Algorithm with an Improved YOLOv5m Network Model

To address the challenges of high complexity,poor real-time performance,and low detection rates for small target vehicles in existing vehicle object detection algorithms,this paper proposes a real-time lightweight architecture based on You Only Look Once(YOLO)v5m.Firstly,a lightweight upsampling operator called Content-Aware Reassembly of Features(CARAFE)is introduced in the feature fusion layer of the network to maximize the extraction of deep-level features for small target vehicles,reducing the missed detection rate and false detection rate.Secondly,a new prediction layer for tiny targets is added,and the feature fusion network is redesigned to enhance the detection capability for small targets.Finally,this paper applies L1 regularization to train the improved network,followed by pruning and fine-tuning operations to remove redundant channels,reducing computational and parameter complexity and enhancing the detection efficiency of the network.Training is conducted on the VisDrone2019-DET dataset.The experimental results show that the proposed algorithmreduces parameters and computation by 63.8% and 65.8%,respectively.The average detection accuracy improves by 5.15%,and the detection speed reaches 47 images per second,satisfying real-time requirements.Compared with existing approaches,including YOLOv5m and classical vehicle detection algorithms,our method achieves higher accuracy and faster speed for real-time detection of small target vehicles in edge computing.

Effective Model Compression via Stage-wise Pruning

Automated machine learning(AutoML)pruning methods aim at searching for a pruning strategy automatically to reduce the computational complexity of deep convolutional neural networks(deep CNNs).However,some previous work found that the results of many Auto-ML pruning methods cannot even surpass the results of the uniformly pruning method.In this paper,the ineffectiveness of Auto-ML pruning,which is caused by unfull and unfair training of the supernet,is shown.A deep supernet suffers from unfull training because it contains too many candidates.To overcome the unfull training,a stage-wise pruning(SWP)method is proposed,which splits a deep supernet into several stage-wise supernets to reduce the candidate number and utilize inplace distillation to supervise the stage training.Besides,a wide supernet is hit by unfair training since the sampling probability of each channel is unequal.Therefore,the fullnet and the tinynet are sampled in each training iteration to ensure that each channel can be overtrained.Remarkably,the proxy performance of the subnets trained with SWP is closer to the actual performance than that of most of the previous AutoML pruning work.Furthermore,experiments show that SWP achieves the state-of-the-art in both CIFAR-10 and ImageNet under the mobile setting.

Dynamic detection method for falling ears of maize harvester based on improved YOLO-V4

Traditional maize ear harvesters mainly rely on manual identification of fallen maize ears,which cannot realize real-time detection of ear falling.The improved You Only Look Once-V4(YOLO-V4)algorithm is combined with the channel pruning algorithm to detect the dropped ears of maize harvesters.K-means clustering algorithm is used to obtain a prior box matching the size of the dropped ears,which improves the Intersection Over Union(IOU).Compare the effect of different activation functions on the accuracy of the YOLO-V4 model,and use the Mish activation function as the activation function of this model.Improve the calculation of the regression positioning loss function,and use the CEIOU loss function to balance the accuracy of each category.Use improved Adam optimization function and multi-stage learning optimization technology to improve the accuracy of the YOLO-V4 model.The channel pruning algorithm is used to compress the model and distillation technology is used in the fine-tuning of the model.The final model size was only 10.77%before compression,and the test set mean Average Precision(mAP)was 93.14%.The detection speed was 112 fps,which can meet the need for real-time detection of maize harvester ears in the field.This study can provide technical reference for the detection of the ear loss rate of intelligent maize harvesters.