Patent
Processing method for convolutional neural network, processing device for convolutional neural network and electronic apparatuses
Li Quanquan,Yan Junjie +1 more
- 29 Jun 2018
5
TL;DR: In this paper, the authors provide a processing method for a convolutional neural network, a processing device for a CNN, and electronic apparatuses, which can be applied to resource-limited equipment to realize the same function.
read more
Abstract: The embodiments of the present invention provide a processing method for a convolutional neural network, a processing device for a convolutional neural network and electronic apparatuses. The processing method for the convolutional neural network includes the following steps that: a trained first convolutional neural network is simplified, so that a second convolutional neural network can be obtained; first training sample data are detected through the first convolutional neural network and the second convolutional neural network, so that first detection data outputted by the first convolutional neural network and second detection data outputted by the second convolutional neural network are obtained; a first difference between the first detection data and the second detection data is determined; and the network parameters of the second convolutional neural network are adjusted according to the first difference at least. With the method and the device of the invention adopted, a small-scale convolutional neural network of which the accuracy can be kept to be not damaged relatively can be obtained; the trained small-scale convolutional neural network can be applied to resource-limited equipment which is low in computing power to realize the same function, and the flexibility of the application of the convolutional neural network can be improved.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Citations
Patent
Neural network training method, image processing method and device, equipment and medium
Peng Baoyun,Jin Xiao +1 more
- 24 May 2019
TL;DR: In this article, the first correlation degree between at least two samples in a sample image set through a first neural network and training a second neural network based on the processing result of the first neural networks on theat least two images and the first relevancy between the at least 2 samples to obtain a target neural network, the network scale of the one neural network being greater than the one of the other neural network.
5
Patent
Image information optimization method and device, electronic equipment and storage medium
Ren Sijie,Chen Yan,Cheng Xuanye,Sun Wenxiu +3 more
- 19 Apr 2019
TL;DR: In this article, the authors proposed an image information optimization method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a plurality of original images with a low signal-to-noise ratio, enabling the original images to be images for the same object, and enabling the feature information included in each original image to be different.
4
Patent
Convolution neural network migration method, device, electronic device and storage medium
Qijie Wei,Hao Wang,Dayong Ding +2 more
- 28 Dec 2018
TL;DR: In this article, a convolution neural network migration method is proposed to improve the last pooling layer of the first convolutional neural network to obtain a second convolution network.
2
Patent
Signal modulation mode identification method and device
Huang Sai,Feng Zhiyong,Chai Lu,Ning Fan +3 more
- 08 Oct 2019
TL;DR: In this article, the authors proposed a signal modulation mode identification method and device, which comprises the steps of acquiring a signal to be identified, generating a signal constellation diagram for the to-be-identified signal, and inputting the signal constellation diagrams and the signal enhancement density constellation diagram into a pre-trained signal recognition neural network to obtain a signal modality of the to be-recognized signal.
2
Patent
Network model processing method and device, electronic equipment and storage medium
Guo Shaopeng,Wang Yujie,Li Quanquan +2 more
- 20 Nov 2020
TL;DR: In this paper, a network model processing method for first-network models is presented, where the first network model comprises N stages, N is an integer greater than or equal to 1, and the N stages have corresponding first structures.
References
•Proceedings Article
Do Deep Nets Really Need to be Deep
Jimmy Ba,Rich Caruana +1 more
- 08 Dec 2014
TL;DR: This paper empirically demonstrate that shallow feed-forward nets can learn the complex functions previously learned by deep nets and achieve accuracies previously only achievable with deep models.
Patent
Model compression and fine-tuning
Sreekanta Reddy Annapureddy Venkata,Daniel Hendricus Franciscus Dijkman,David Jonathan Julian +2 more
- 04 Sep 2015
TL;DR: In this article, a neural network is decomposed by replacing one layer in the neural network with compressed layers to produce the compressed network, which can be fine-tuned by updating weight values in the compressed layer(s).
50
Patent
Random convolutional neural network-based high-resolution image scene classification method
Bo Du,Fan Zhang,Liangpei Zhang +2 more
- 21 Dec 2016
TL;DR: In this article, a random convolutional neural network-based high-resolution image scene classification method is proposed, where features are hierarchically learned by using a deep convolution network, and model aggregation learning is carried out by utilizing a gradient upgrading method, so that the problem that a single model easily falls into a local optimal solution is solved and the network generalization capability is improved.
23
Patent
Face recognition method and system
Gong Xuchao
- 07 Dec 2016
TL;DR: In this paper, a face recognition method and system is described, which consists of the steps that an acquired face image under recognition is inputted to a first deep convolution neural network so as to determine the probability that the face image is from someone, wherein the first deep CNN was determined based on deep learning of sample face images; when the probability is greater than a preset threshold, the face under recognition was ensured to be from someone.
18
Patent
Method and system for achieving classification of pedestrians and vehicles based on neural network
Bin Leng,He Qing,Guan Guan,Hu Huan,Jiang Dongguo +4 more
- 08 Apr 2015
TL;DR: In this paper, a method and a system for achieving the classification of pedestrians and vehicles based on a convolutional neural network (CNN) was proposed. And the method includes the following steps of collecting a plurality of training samples, classifying the training samples by a CNN, and thereby obtaining a classifier including tab results, when the pedestrians and the vehicles are classified, reading a video image to be detected, detecting moving objects in the image, and processing the image in blocks according to the moving objects; and then classifying image blocks by the classifier to obtain a detection
13
Related Papers (5)
Fei Rong,Li Shasha,Xu Qingzheng,Liu Kun +3 more
- 26 Nov 2018
Liu Jiwei,Tieyi Wang,Sanyi Yuan +2 more
- 11 Dec 2018