Quantized Convolutional Neural Networks for Mobile Devices

Wu, Jiaxiang; Leng, Cong; Wang, Yuhang; Hu, Qinghao; Cheng, Jian

doi:10.1109/cvpr.2016.521

Cited by 980 publications

(580 citation statements)

References 19 publications

Supporting

Mentioning

574

Contrasting

Unclassified

Order By: Relevance

“…To tackle the storage issue of deep networks, Gong et al [6], Wu et al [31] and Lin et al [18] consider applying the quantization techniques to pre-trained CNNs, and trying to make network compressions with minor concessions on the inference accuracy. Another powerful category of methods in this scope is network pruning.…”

mentioning

confidence: 99%

Network Sketching: Exploiting Binary Structure in Deep CNNs

Guo¹,

Yao²,

Zhao

et al. 2017

2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

100

View full text Add to dashboard Cite

Convolutional neural networks (CNNs) with deep architectures have substantially advanced the state-of-the-art in computer vision tasks. However, deep networks are typically resource-intensive and thus difficult to be deployed on mobile devices. Recently, CNNs with binary weights have shown compelling efficiency to the community, whereas the accuracy of such models is usually unsatisfactory in practice. In this paper, we introduce network sketching as a novel technique of pursuing binary-weight CNNs, targeting at more faithful inference and better trade-off for practical applications. Our basic idea is to exploit binary structure directly in pre-trained filter banks and produce binaryweight models via tensor expansion. The whole process can be treated as a coarse-to-fine model approximation, akin to the pencil drawing steps of outlining and shading. To further speedup the generated models, namely the sketches, we also propose an associative implementation of binary tensor convolutions. Experimental results demonstrate that a proper sketch of AlexNet (or ResNet) outperforms the existing binary-weight models by large margins on the ImageNet large scale classification task, while the committed memory for network parameters only exceeds a little.

show abstract

mentioning

confidence: 99%

Network Sketching: Exploiting Binary Structure in Deep CNNs

Guo¹,

Yao²,

Zhao

et al. 2017

2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

100

View full text Add to dashboard Cite

show abstract

“…Therefore, we introduce our approach on adapting SqueezeNet (Iandola et al, 2016), a smaller CNN with a model size of only 4,8 MB which is even 10 times smaller than GoogLeNet. In future work this might be reduced even more by quantizing (Wu et al, 2016) or binarization (Courbariaux et al, 2016).…”

Section: Related Workmentioning

confidence: 99%

Squeezeposenet: Image Based Pose Regression With Small Convolutional Neural Networks for Real Time Uas Navigation

Müller

Urban

Jutzi

2017

ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

ABSTRACT:The number of unmanned aerial vehicles (UAVs) is increasing since low-cost airborne systems are available for a wide range of users. The outdoor navigation of such vehicles is mostly based on global navigation satellite system (GNSS) methods to gain the vehicles trajectory. The drawback of satellite-based navigation are failures caused by occlusions and multi-path interferences. Beside this, local image-based solutions like Simultaneous Localization and Mapping (SLAM) and Visual Odometry (VO) can e.g. be used to support the GNSS solution by closing trajectory gaps but are computationally expensive. However, if the trajectory estimation is interrupted or not available a re-localization is mandatory. In this paper we will provide a novel method for a GNSS-free and fast image-based pose regression in a known area by utilizing a small convolutional neural network (CNN). With on-board processing in mind, we employ a lightweight CNN called SqueezeNet and use transfer learning to adapt the network to pose regression. Our experiments show promising results for GNSS-free and fast localization.

show abstract

“…Such techniques focus on compressing an already trained CNN, instead of training a CNN with fewer parameters in the first place. Some of these works [7], [9], also use vector quantization techniques. However, proposed the method uses a differentiable quantization scheme that allows for training both the quantizer and the rest of the network simultaneously.…”

Section: Related Workmentioning

confidence: 99%

“…Many techniques have been proposed to reduce the model size [6]- [9]. Usually compression and pruning techniques are used to reduce the size of CNN models [6], [7], [9]. Such techniques focus on compressing an already trained CNN, instead of training a CNN with fewer parameters in the first place.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Concept detection and face pose estimation using lightweight convolutional neural networks for steering drone video shooting

Passalis

Tefas

2017

2017 25th European Signal Processing Conference (EUSIPCO)

View full text Add to dashboard Cite

Abstract-Unmanned Aerial Vehicles, also known as drones, are becoming increasingly popular for video shooting tasks since they are capable of capturing spectacular aerial shots. Deep learning techniques, such as Convolutional Neural Networks (CNNs), can be utilized to assist various aspects of the flying and the shooting process allowing one human to operate one or more drones at once. However, using deep learning techniques on drones is not straightforward since computational power and memory constraints exist. In this work, a quantizationbased method for learning lightweight convolutional networks is proposed. The ability of the proposed approach to significantly reduce the model size and increase both the feed-forward speed and the accuracy is demonstrated on two different drone-related tasks, i.e., human concept detection and face pose estimation.

show abstract

Quantized Convolutional Neural Networks for Mobile Devices

Cited by 980 publications

References 19 publications

Network Sketching: Exploiting Binary Structure in Deep CNNs

Network Sketching: Exploiting Binary Structure in Deep CNNs

Squeezeposenet: Image Based Pose Regression With Small Convolutional Neural Networks for Real Time Uas Navigation

Concept detection and face pose estimation using lightweight convolutional neural networks for steering drone video shooting

Contact Info

Product

Resources

About