In recent years, deep neural networks have been successful in both industry and academia, especially for computer vision tasks. The great success of deep learning is mainly due to its scalability to encode large-scale data and to maneuver billions of model parameters. However, it is a challenge to deploy these cumbersome deep models on devices with limited resources, e.g., mobile phones and embedded devices, not only because of the high computational complexity but also the large storage requirements. To this end, a variety of model compression and acceleration techniques have been developed. As a representative type of model compression and acceleration, knowledge distillation effectively learns a small student model from a large teacher model. It has received rapid increasing attention from the community. This paper provides a comprehensive survey of knowledge distillation from the perspectives of knowledge categories, training schemes, teacher-student architecture, distillation algorithms, performance comparison and applications. Furthermore, challenges in knowledge distillation are briefly reviewed and comments on future research are discussed and forwarded.

Knowledge Distillation: A Survey

Automatic segmentation of retinal vessels in fundus images plays an important role in the diagnosis of some diseases such as diabetes and hypertension. In this paper, we propose Deformable U-Net (DUNet), which exploits the retinal vessels’ local features with a U-shape architecture, in an end to end manner for retinal vessel segmentation. Inspired by the recently introduced deformable convolutional networks, we integrate the deformable convolution into the proposed network. The DUNet, with upsampling operators to increase the output resolution, is designed to extract context information and enable precise localization by combining low-level features with high-level ones. Furthermore, DUNet captures the retinal vessels at various shapes and scales by adaptively adjusting the receptive fields according to vessels’ scales and shapes. Public datasets: DRIVE, STARE, CHASE_DB1 and HRF are used to test our models. Detailed comparisons between the proposed network and the deformable neural network, U-Net are provided in our study. Results show that more detailed vessels can be extracted by DUNet and it exhibits state-of-the-art performance for retinal vessel segmentation with a global accuracy of 0.9566/0.9641/0.9610/0.9651 and AUC of 0.9802/0.9832/0.9804/0.9831 on DRIVE, STARE, CHASE_DB1 and HRF respectively. Moreover, to show the generalization ability of the DUNet, we use another two retinal vessel data sets, i.e., WIDE and SYNTHE, to qualitatively and quantitatively analyze and compare with other methods. Extensive cross-training evaluations are used to further assess the extendibility of DUNet. The proposed method has the potential to be applied to the early diagnosis of diseases.

DUNet: A deformable network for retinal vessel segmentation

Deep neural models, in recent years, have been successful in almost every field. However, these models are huge, demanding heavy computation power. Besides, the performance boost is highly dependent on redundant labeled data. To achieve faster speeds and to handle the problems caused by the lack of labeled data, knowledge distillation (KD) has been proposed to transfer information learned from one model to another. KD is often characterized by the so-called ‘Student-Teacher’ (S-T) learning framework and has been broadly applied in model compression and knowledge transfer. This paper is about KD and S-T learning, which are being actively studied in recent years. First, we aim to provide explanations of what KD is and how/why it works. Then, we provide a comprehensive survey on the recent progress of KD methods together with S-T frameworks typically used for vision tasks. In general, we investigate some fundamental questions that have been driving this research area and thoroughly generalize the research progress and technical details. Additionally, we systematically analyze the research status of KD in vision applications. Finally, we discuss the potentials and open challenges of existing methods and prospect the future directions of KD and S-T learning.

Knowledge Distillation and Student-Teacher Learning for Visual Intelligence: A Review and New Outlooks.

Diabetic retinopathy detection through deep learning techniques: A review

Artificial intelligence (AI) has emerged as a major frontier in computer science research. Although AI has broad application across many medical fields, it will have particular utility in ophthalmology and will dramatically change the diagnostic and treatment pathways for many eye conditions such as corneal ectasias, glaucoma, age-related macular degeneration and diabetic retinopathy. However, given that AI has primarily been driven as a computer science, its concepts and terminology are unfamiliar to many medical professionals. Important key terms such as machine learning and deep learning are often misunderstood and incorrectly used interchangeably. This article presents an overview of AI and new developments relevant to ophthalmology.

/pdf/current-state-and-future-prospects-of-artificial-42r84iqkyw.pdf

Current state and future prospects of artificial intelligence in ophthalmology: a review.

The automatic detection of diabetic retinopathy is of vital importance, as it is the main cause of irreversible vision loss in the working-age population in the developed world. The early detection of diabetic retinopathy occurrence can be very helpful for clinical treatment; although several different feature extraction approaches have been proposed, the classification task for retinal images is still tedious even for those trained clinicians. Recently, deep convolutional neural networks have manifested superior performance in image classification compared to previous handcrafted feature-based image classification methods. Thus, in this paper, we explored the use of deep convolutional neural network methodology for the automatic classification of diabetic retinopathy using color fundus image, and obtained an accuracy of 94.5% on our dataset, outperforming the results obtained by using classical approaches.

https://www.mdpi.com/1420-3049/22/12/2054/pdf

Deep Convolutional Neural Network-Based Early Automated Detection of Diabetic Retinopathy Using Fundus Image.

The audio classification task aims to discriminate between different audio signal types. In this task, deep neural networks have achieved better performance than the traditional shallow architecture-based machine-learning method. However, deep neural networks often require huge computational and storage requirements that hinder the deployment in embedded devices. In this paper, we proposed a distillation method which transfers knowledge from well-trained networks to a small network, and the method can compress model size while improving audio classification precision. The contributions of the proposed method are two folds: a multi-level feature distillation method was proposed and an adversarial learning strategy was employed to improve the knowledge transfer. The extensive experiments are conducted on three audio classification tasks, audio scene classification, general audio tagging, and speech command recognition. The experimental results demonstrate that: the small network can provide better performance while achieves the calculated amount of floating-point operations per second (FLOPS) compression ratio of 76:1 and parameters compression ratio of 3:1.

/pdf/an-adversarial-feature-distillation-method-for-audio-2082r74eho.pdf

An Adversarial Feature Distillation Method for Audio Classification

Audio tagging aims to infer descriptive labels from audio clips and it is challenging due to the limited size of data and noisy labels. The solution to the tagging task is described in this paper. The main contributions include the following: an ensemble learning framework is applied to ensemble statistical features and the outputs from the deep classifiers, with the goal to utilize complementary information. Moreover, a sample re-weight strategy is employed to address the noisy label problem within the framework. The approach achieves a mean average precision of 0.958, outperforming the baseline system with a large margin.

/pdf/general-audio-tagging-with-ensembling-convolutional-neural-4deg6yxytw.pdf

General audio tagging with ensembling convolutional neural networks and statistical features.

B-mode ultrasound tongue imaging is widely used in the speech production field. However, efficient interpretation is in a great need for the tongue image sequences. Inspired by the recent success of unsupervised deep learning approach, we explore unsupervised convolutional network architecture for the feature extraction in the ultrasound tongue image, which can be helpful for the clinical linguist and phonetics. By quantitative comparison between different unsupervised feature extraction approaches, the denoising convolutional autoencoder (DCAE)-based method outperforms the other feature extraction methods on the reconstruction task and the 2010 silent speech interface challenge. A Word Error Rate of 6.17% is obtained with DCAE, compared to the state-of-the-art value of 6.45% using Discrete cosine transform as the feature extractor. Our codes are available at https://github.com/DeePBluE666/Source-code1.

Denoising Convolutional Autoencoder Based B-mode Ultrasound Tongue Image Feature Extraction

Recently, deep learning has achieved state-of-the-art performance in more aspects than traditional shallow architecture-based machine-learning methods. However, in order to achieve higher accuracy, it is usually necessary to extend the network depth or ensemble the results of different neural networks. Increasing network depth or ensembling different networks increases the demand for memory resources and computing resources. This leads to difficulties in deploying depth-learning models in resource-constrained scenarios such as drones, mobile phones, and autonomous driving. Improving network performance without expanding the network scale has become a hot topic for research. In this paper, we propose a cross-architecture online-distillation approach to solve this problem by transmitting supplementary information on different networks. We use the ensemble method to aggregate networks of different structures, thus forming better teachers than traditional distillation methods. In addition, discontinuous distillation with progressively enhanced constraints is used to replace fixed distillation in order to reduce loss of information diversity in the distillation process. Our training method improves the distillation effect and achieves strong network-performance improvement. We used some popular models to validate the results. On the CIFAR100 dataset, AlexNet's accuracy was improved by 5.94%, VGG by 2.88%, ResNet by 5.07%, and DenseNet by 1.28%. Extensive experiments were conducted to demonstrate the effectiveness of the proposed method. On the CIFAR10, CIFAR100, and ImageNet datasets, we observed significant improvements over traditional knowledge distillation.

https://www.mdpi.com/1099-4300/21/4/357/pdf

Haibo Mi

Papers

Deep Convolutional Neural Network-Based Early Automated Detection of Diabetic Retinopathy Using Fundus Image.

An Adversarial Feature Distillation Method for Audio Classification

General audio tagging with ensembling convolutional neural networks and statistical features.

Denoising Convolutional Autoencoder Based B-mode Ultrasound Tongue Image Feature Extraction

Multistructure-Based Collaborative Online Distillation