A Compact DNN: Approaching GoogLeNet-Level Accuracy of Classification and Domain Adaptation

Wu, Chunpeng; Wen, Wei; Afzal, Tariq; Zhang, Yongmei; Chen, Yiran; Li, Hai

doi:10.1109/cvpr.2017.88

Cited by 49 publications

(20 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Interestingly, many compact deep neural network (DNN) models can still achieve an excellent performance on image classification, such as SquezeNet [21], Compact DNN [22], MobileNetV1 [23], ShuffleNet [24], and MobileNetV2 [25]. Some popular applications of object recognition include gesture recognition [26], handwritten Chinese text recognition [27] and traffic sign classification [28].…”

Section: Compact Designmentioning

confidence: 99%

A Compact Convolutional Neural Network for Surface Defect Inspection

Huang

Qiu

Wang

et al. 2020

Sensors

View full text Add to dashboard Cite

The advent of convolutional neural networks (CNNs) has accelerated the progress of computer vision from many aspects. However, the majority of the existing CNNs heavily rely on expensive GPUs (graphics processing units). to support large computations. Therefore, CNNs have not been widely used to inspect surface defects in the manufacturing field yet. In this paper, we develop a compact CNN-based model that not only achieves high performance on tiny defect inspection but can be run on low-frequency CPUs (central processing units). Our model consists of a light-weight (LW) bottleneck and a decoder. By a pyramid of lightweight kernels, the LW bottleneck provides rich features with less computational cost. The decoder is also built in a lightweight way, which consists of an atrous spatial pyramid pooling (ASPP) and depthwise separable convolution layers. These lightweight designs reduce the redundant weights and computation greatly. We train our models on groups of surface datasets. The model can successfully classify/segment surface defects with an Intel i3-4010U CPU within 30 ms. Our model obtains similar accuracy with MobileNetV2 while only has less than its 1/3 FLOPs (floating-point operations per second) and 1/8 weights. Our experiments indicate CNNs can be compact and hardware-friendly for future applications in the automated surface inspection (ASI). pixels into regions, with the intention to emphasize the regions corresponding to surface defects. For example, Reference [6] firstly preprocesses images using convex optimization, then segments wood defects with Otsu segmentation method. Reference [12] segments fabric defects by the Sobel and the watershed algorithm.These traditional methods involve image preprocessing, feature extraction, feature reduction, and classifier selection, which needs the experience of experts. Besides, most surface defects are tiny and very similar to their background, even the inspection for a single type of defect can be very difficult for these traditional methods. Although traditional algorithms are designed for specific surfaces, effective measures such as recall, precision, and accuracy can barely achieve the standard of industrial applications. This long-existing bottleneck has been restraining the growth of ASI.Convolutional neural networks (CNNs) greatly stimulate the development of ASI [3,13-15] as deep learning methods in computer vision are achieving the state-of-the-art in recent years. As end-to-end solutions, CNNs also simplify the procedures of ASI. For example, a fully convolutional network (FCN) can segment surface defects by supervised learning [16], without any extra pre-process or post-process.A notorious limitation of CNNs is computation. In many cases, expensive GPUs with high computing capabilities are necessary to support large workloads. However, in considering manufacturing cost, most core computing resources of ASI platforms are mainly the low-power CPUs of industrial personal computers (IPC), or even FPGAs (field programmable gate array). To take advantage...

show abstract

Section: Compact Designmentioning

confidence: 99%

A Compact Convolutional Neural Network for Surface Defect Inspection

Huang

Qiu

Wang

et al. 2020

Sensors

View full text Add to dashboard Cite

show abstract

“…Various CNNs trained models are studied with different imaging applications [2]. The AlexNet architecture is applied that proved CNNs performed better because of less number training samples.…”

Section: Literature Reviewmentioning

confidence: 99%

Convolutional Neural Network Based Medical Image Classifier

Sundararajan*¹,

Sivagurunathan²,

Venkatesh³

et al. 2019

IJRTE

View full text Add to dashboard Cite

Deep learning had provided good outcome in analyzing images of tumours, however, the deficiency of large annotated datasets reduces its importance. The proposed medical image processing system is based on image segmentation and image classification. It is to be used by medical field experts. In order to classify the Brain tumour images the semantic level classification and segmentation network techniques are applied. This includes prior knowledge of testing samples and training samples, using Convolutional Neural Networks (CNN). The CNN based classifier improves the detection accuracy compared to the existing segmentation based classifier. In this project, the automated system would help the medical image analyst to identify the Brain Tumour in patient by making use of deep convolutional neural network (CNN).The image is obtained from MRI scan of a brain. The tumourless patient’s image dataset is used as the training and testing data for the classification network. Patient image is compared with dataset of a tumour affected images for differentiating an image as Non-timorous sample, low grade glioma and glioblastoma after segmenting and classifying the image. The Watershed segmentation algorithm is used for segmenting images and CNN is used for classifying the images. Finally the system will detect the tumour is affected or not in the given image of a patient’s brain, then the system will identify the tumour affected region and differentiate the low grade glioma and glioblastoma in the image.

show abstract

“…Accuracy Top-1 Size(MB) AlexNet [16] 57.1 15.0 VGG-16 [24] 71.5 58.9 Googlenet [25] 69.8 23.2 ShuffleNet-2x [33] 70.9 19.2 1.0 MobileNet-224 [12] 70.6 15.9 Compact DNN [28] 68.9 13.6 Gnet-1(ours) 67.0 5.5 Gnet-2(ours) 58.1 2.8 convolutional architecture as VGG, and the FC layers are deployed outside the accelerator. In Section 6, we will show that FC can be implemented within our CNN-DSA accelerator as well.…”

Section: Modelmentioning

confidence: 99%

Ultra Power-Efficient CNN Domain Specific Accelerator with 9.3TOPS/Watt for Mobile and Embedded Applications

Sun¹,

Yang²,

Dong³

et al. 2018

2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

Computer vision performances have been significantly improved in recent years by Convolutional Neural Networks (CNN). Currently, applications using CNN algorithms are deployed mainly on general purpose hardwares, such as CPUs, GPUs or FPGAs. However, power consumption, speed, accuracy, memory footprint, and die size should all be taken into consideration for mobile and embedded applications. Domain Specific Architecture (DSA) for CNN is the efficient and practical solution for CNN deployment and implementation. We designed and produced a 28nm Two-Dimensional CNN-DSA accelerator with an ultra powerefficient performance of 9.3TOPS/Watt and with all processing done in the internal memory instead of outside DRAM. It classifies 224x224 RGB image inputs at more than 140fps with peak power consumption at less than 300mW and an accuracy comparable to the VGG benchmark. The CNN-DSA accelerator is reconfigurable to support CNN model coefficients of various layer sizes and layer types, including convolution, depth-wise convolution, short-cut connections, max pooling, and ReLU. Furthermore, in order to better support real-world deployment for various application scenarios, especially with low-end mobile and embedded platforms and MCUs (Microcontroller Units), we also designed algorithms to fully utilize the CNN-DSA accelerator efficiently by reducing the dependency on external accelerator computation resources, including implementation of Fully-Connected (FC) layers within the accelerator and compression of extracted features from the CNN-DSA accelerator. Live demos with our CNN-DSA accelerator on mobile and embedded systems show its capabilities to be widely and practically applied in the real world.

show abstract

A Compact DNN: Approaching GoogLeNet-Level Accuracy of Classification and Domain Adaptation

Cited by 49 publications

References 26 publications

A Compact Convolutional Neural Network for Surface Defect Inspection

A Compact Convolutional Neural Network for Surface Defect Inspection

Convolutional Neural Network Based Medical Image Classifier

Ultra Power-Efficient CNN Domain Specific Accelerator with 9.3TOPS/Watt for Mobile and Embedded Applications

Contact Info

Product

Resources

About