This paper tackles the problem of training a deep convolutional neural network with both low-precision weights and low-bitwidth activations. Optimizing a low-precision network is very challenging since the training process can easily get trapped in a poor local minima, which results in substantial accuracy loss. To mitigate this problem, we propose three simple-yet-effective approaches to improve the network training. First, we propose to use a two-stage optimization strategy to progressively find good local minima. Specifically, we propose to first optimize a net with quantized weights and then quantized activations. This is in contrast to the traditional methods which optimize them simultaneously. Second, following a similar spirit of the first method, we propose another progressive optimization approach which progressively decreases the bit-width from high-precision to low-precision during the course of training. Third, we adopt a novel learning scheme to jointly train a full-precision model alongside the low-precision one. By doing so, the full-precision model provides hints to guide the low-precision model training. Extensive experiments on various datasets ( i.e., CIFAR-100 and ImageNet) show the effectiveness of the proposed methods. To highlight, using our methods to train a 4-bit precision network leads to no performance decrease in comparison with its full-precision counterpart with standard network architectures ( i.e., AlexNet and ResNet-50).
University of Washington（华盛顿大学）；Microsoft Research（微软研究院）
Recent work has shown that fast, compact low-bitwidth neural networks can be surprisingly accurate. These networks use homogeneous binarization: all parameters in each layer or (more commonly) the whole model have the same low bitwidth (e.g., 2 bits). However, modern hardware allows efficient designs where each arithmetic instruction can have a custom bitwidth, motivating heterogeneous binarization, where every parameter in the network may have a different bitwidth. In this paper, we show that it is feasible and useful to select bitwidths at the parameter granularity during training. For insta…
Monash University（莫纳什大学）；The University of Adelaide（阿德莱德大学）；South China University of Technology（华南理工大学）
This paper tackles the problem of training a deep convolutional neural network of both low-bitwidth weights and activations. Optimizing a low-precision network is very challenging due to the non-differentiability of the quantizer, which may result in substantial accuracy loss. To address this, we propose three practical approaches, including (i) progressive quantization; (ii) stochastic precision; and (iii) joint knowledge distillation to improve the network training. First, for progressive quantization, we propose two schemes to progressively find good local minima. Specifically, we propose t…
Efficient model inference is an important and practical issue in the deployment of deep neural network on resource constraint platforms. Network quantization addresses this problem effectively by leveraging low-bit representation and arithmetic that could be conducted on dedicated embedded systems. In the previous works, the parameter bitwidth is set homogeneously and there is a trade-off between superior performance and aggressive compression. Actually the stacked network layers, which are generally regarded as hierarchical feature extractors, contribute diversely to the overall performance. …
Deep neural network quantization with adaptive bitwidths has gained increasing attention due to the ease of model deployment on various platforms with different resource budgets. In this paper, we propose a meta-learning approach to achieve this goal. Specifically, we propose MEBQAT, a simple yet effective way of bitwidth-adaptive quantization aware training (QAT) where meta-learning is effectively combined with QAT by redefining meta-learning tasks to incorporate bitwidths. After being deployed on a platform, MEBQAT allows the (meta-)trained model to be quantized to any candidate bitwidth the…
The University of Hong Kong（香港大学）；SUSTech（南方科技大学）；The Chinese University of Hong Kong（香港中文大学）
Learning convolutional neural networks (CNNs) with low bitwidth is challenging because performance may drop significantly after quantization. Prior arts often discretize the network weights by carefully tuning hyper-parameters of quantization (e.g. non-uniform stepsize and layer-wise bitwidths), which are complicated and sub-optimal because the full-precision and low-precision models have a large discrepancy. This work presents a novel quantization pipeline, Frequency-Aware Transformation (FAT), which has several appealing benefits. (1) Rather than designing complicated quantizers like existin…
In recent years, deep neural network exhibits its powerful superiority on information discrimination in many computer vision applications. However, the capacity of deep neural network architecture is still a mystery to the researchers. Intuitively, larger capacity of neural network can always deposit more information to improve the discrimination ability of the model. But, the learnable parameter scale is not feasible to estimate the capacity of deep neural network. Due to the overfitting, directly increasing hidden nodes number and hidden layer number are already demonstrated not necessary to…
South China University of Technology（华南理工大学）；PengCheng Laboratory（鹏城实验室）；Monash University（蒙纳士大学）
Neural network quantization is an effective way to compress deep models and improve their execution latency and energy efficiency, so that they can be deployed on mobile or embedded devices. Existing quantization methods require original data for calibration or fine-tuning to get better performance. However, in many real-world scenarios, the data may not be available due to confidential or private issues, thereby making existing quantization methods not applicable. Moreover, due to the absence of original data, the recently developed generative adversarial networks (GANs) cannot be applied to …
Zhejiang University（浙江大学）；China Research and Development Academy of Machinery Equipment（中国机械科学研究总院集团有限公司）
The quantized neural networks (QNNs) can be useful for neural network acceleration and compression, but during the training process they pose a challenge: how to propagate the gradient of loss function through the graph flow with a derivative of 0 almost everywhere. In response to this non-differentiable situation, we propose a novel Asymptotic-Quantized Estimator (AQE) to estimate the gradient. In particular, during back-propagation, the graph that relates inputs to output remains smoothness and differentiability. At the end of training, the weights and activations have been quantized to low-…
BitQ: Tailoring Block Floating Point Precision for Improved DNN Efficiency on Resource-Constrained Devices
Guangdong University of Technology（广东工业大学）；Beijing Normal University - Hong Kong Baptist University United International College（北京师范大学-香港浸会大学联合国际学院）；Lanzhou University（兰州大学）；Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）
Hong Kong University of Science and Technology（香港科技大学）；Mohamed bin Zayed University of Artificial Intelligence（穆罕默德·本·扎耶德人工智能大学）；Jockey Club Institute for Advanced Study, HKUST（香港科技大学赛马会高等研究院）；Reality Labs, Meta Inc（Meta公司现实实验室）；ACCESS - AI Chip Center for Emerging Smart Systems（新兴智能系统AI芯片中心（ACCESS））；Carnegie Mellon University（卡内基梅隆大学）
In order to deploy deep models in a computationally efficient manner, model quantization approaches have been frequently used. In addition, as new hardware that supports mixed bitwidth arithmetic operations, recent research on mixed precision quantization (MPQ) begins to fully leverage the capacity of representation by searching optimized bitwidths for different layers and modules in a network. However, previous studies mainly search the MPQ strategy in a costly scheme using reinforcement learning, neural architecture search, etc., or simply utilize partial prior knowledge for bitwidth assignm…
Eindhoven University of Technology（埃因霍温理工大学）；Philips Innovation Services（飞利浦创新服务）
Convolutional Neural Networks (CNNs) have proven to be a powerful state-of-the-art method for image classification tasks. One drawback however is the high computational complexity and high memory consumption of CNNs which makes them unfeasible for execution on embedded platforms which are constrained on physical resources needed to support CNNs. Quantization has often been used to efficiently optimize CNNs for memory and computational complexity at the cost of a loss of prediction accuracy. We therefore propose a method to optimally quantize the weights, biases and activations of each layer of…
Efficient deep neural network (DNN) inference on mobile or embedded devices typically involves quantization of the network parameters and activations. In particular, mixed precision networks achieve better performance than networks with homogeneous bitwidth for the same size constraint. Since choosing the optimal bitwidths is not straight forward, training methods, which can learn them, are desirable. Differentiable quantization with straight-through gradients allows to learn the quantizer's parameters using gradient methods. We show that a suited parametrization of the quantizer is the key to…
We present PrecisionBatching, a quantized inference algorithm for speeding up neural network execution on traditional hardware platforms at low bitwidths without the need for retraining or recalibration. PrecisionBatching decomposes a neural network into individual bitlayers and accumulates them using fast 1-bit operations while maintaining activations in full precision. PrecisionBatching not only facilitates quantized inference at low bitwidths (< 8 bits) without the need for retraining/recalibration, but also 1) enables traditional hardware platforms the ability to realize inference speedups…
University of Chinese Academy of Sciences（中国科学院大学）；Megvii Inc.（旷视科技有限公司）
Quantized Neural Networks (QNNs), which use low bitwidth numbers for representing parameters and performing computations, have been proposed to reduce the computation complexity, storage size and memory usage. In QNNs, parameters and activations are uniformly quantized, such that the multiplications and additions can be accelerated by bitwise operations. However, distributions of parameters in Neural Networks are often imbalanced, such that the uniform quantization determined from extremal values may under utilize available bitwidth. In this paper, we propose a novel quantization method that c…
ETH Zurich（苏黎世联邦理工学院）；Singapore Management University（新加坡管理大学）
We investigate the compression of deep neural networks by quantizing their weights and activations into multiple binary bases, known as multi-bit networks (MBNs), which accelerate the inference and reduce the storage for the deployment on low-resource mobile and embedded platforms. We propose Adaptive Loss-aware Quantization (ALQ), a new MBN quantization pipeline that is able to achieve an average bitwidth below one-bit without notable loss in inference accuracy. Unlike previous MBN quantization solutions that train a quantizer by minimizing the error to reconstruct full precision weights, ALQ…
Model quantization is a widely used technique to compress and accelerate deep neural network (DNN) inference. Emergent DNN hardware accelerators begin to support mixed precision (1-8 bits) to further improve the computation efficiency, which raises a great challenge to find the optimal bitwidth for each layer: it requires domain experts to explore the vast design space trading off among accuracy, latency, energy, and model size, which is both time-consuming and sub-optimal. Conventional quantization algorithm ignores the different hardware architectures and quantizes all the layers in a unifor…
Russian Research Center Huawei Technologies, Russia（华为技术有限公司俄罗斯研究中心）；Department of Computer Science ETH Zurich, Switzerland（苏黎世联邦理工学院计算机科学系）
Despite the growing popularity of deep learning technologies, high memory requirements and power consumption are essentially limiting their application in mobile and IoT areas. While binary convolutional networks can alleviate these problems, the limited bitwidth of weights is often leading to significant degradation of prediction accuracy. In this paper, we present a method for training binary networks that maintains a stable predefined level of their information capacity throughout the training process by applying Shannon entropy based penalty to convolutional filters. The results of experim…
In this paper, we propose to train a network with binary weights and low-bitwidth activations, designed especially for mobile devices with limited power consumption. Most previous works on quantizing CNNs uncritically assume the same architecture, though with reduced precision. However, we take the view that for best performance it is possible (and even likely) that a different architecture may be better suited to dealing with low precision weights and activations. Specifically, we propose a "network expansion" strategy in which we aggregate a set of homogeneous low-precision branches to impli…
We present Convolutional Mean (CM) - a simple and fast convolutional neural network for illuminant estimation. Our proposed method only requires a small neural network model (1.1K parameters) and a 48 x 32 thumbnail input image. Our unoptimized Python implementation takes 1 ms/image, which is arguably 3-3750x faster than the current leading solutions with similar accuracy. Using two public datasets, we show that our proposed light-weight method offers accuracy comparable to the current leading methods' (which consist of thousands/millions of parameters) across several measures.
A standard deep convolutional neural network paired with a suitable loss function learns compact local image descriptors that perform comparably to state-of-the art approaches.
Radboud University Medical Center（拉德堡德大学医学中心）；Computational Pathology Group（计算病理学小组）
To train deep convolutional neural networks, the input data and the intermediate activations need to be kept in memory to calculate the gradient descent step. Given the limited memory available in the current generation accelerator cards, this limits the maximum dimensions of the input data. We demonstrate a method to train convolutional neural networks holding only parts of the image in memory while giving equivalent results. We quantitatively compare this new way of training convolutional neural networks with conventional training. In addition, as a proof of concept, we train a convolutional…
Trained on the works of an artist under study and visually comparable works of other artists, convolutional neural networks can identify forgeries and provide attributions. They can also assign classification probabilities within a painting, revealing mixed authorship and identifying regions painted by different hands.
Institution of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）
Convolutional kernels are basic and vital components of deep Convolutional Neural Networks (CNN). In this paper, we equip convolutional kernels with shape attributes to generate the deep Irregular Convolutional Neural Networks (ICNN). Compared to traditional CNN applying regular convolutional kernels like ${3\times3}$, our approach trains irregular kernel shapes to better fit the geometric variations of input features. In other words, shapes are learnable parameters in addition to weights. The kernel shapes and weights are learned simultaneously during end-to-end training with the standard bac…
Politecnico di Milano（米兰理工大学）；University of Montreal（蒙特利尔大学）
In this paper, we propose a deep neural network architecture for object recognition based on recurrent neural networks. The proposed network, called ReNet, replaces the ubiquitous convolution+pooling layer of the deep convolutional neural network with four recurrent neural networks that sweep horizontally and vertically in both directions across the image. We evaluate the proposed ReNet on three widely-used benchmark datasets; MNIST, CIFAR-10 and SVHN. The result suggests that ReNet is a viable alternative to the deep convolutional neural network, and that further investigation is needed.
We present a deep learning approach to the ISIC 2017 Skin Lesion Classification Challenge using a multi-scale convolutional neural network. Our approach utilizes an Inception-v3 network pre-trained on the ImageNet dataset, which is fine-tuned for skin lesion classification using two different scales of input images.
This article presents the design, experiments and results of our solution submitted to the 2018 ISIC challenge: Skin Lesion Analysis Towards Melanoma Detection. We design a pipeline using state-of-the-art Convolutional Neural Network (CNN) models for a Lesion Boundary Segmentation task and a Lesion Diagnosis task.
In this project we work on creating a model to classify images for the Pascal VOC Challenge 2012. We use convolutional neural networks trained on a single GPU instance provided by Amazon via their cloud service Amazon Web Services (AWS) to classify images in the Pascal VOC 2012 data set. We train multiple convolutional neural network models and finally settle on the best model which produced a validation accuracy of 85.6% and a testing accuracy of 85.24%.
We have implemented a convolutional neural network designed for processing sparse three-dimensional input data. The world we live in is three dimensional so there are a large number of potential applications including 3D object recognition and analysis of space-time objects. In the quest for efficiency, we experiment with CNNs on the 2D triangular-lattice and 3D tetrahedral-lattice.
Taras Shevchenko National University of Kyiv（基辅塔拉斯·舍甫琴科国立大学）
At this work we introduce horizontally symmetric convolutional kernels for CNNs which make the network output invariant to horizontal flips of the image. We also study other types of symmetric kernels which lead to vertical flip invariance, and approximate rotational invariance. We show that usage of such kernels acts as regularizer, and improves generalization of the convolutional neural networks at the cost of more complicated training process.
Activation functions play a vital role in the training of Convolutional Neural Networks. For this reason, to develop efficient and performing functions is a crucial problem in the deep learning community. Key to these approaches is to permit a reliable parameter learning, avoiding vanishing gradient problems. The goal of this work is to propose an ensemble of Convolutional Neural Networks trained using several different activation functions. Moreover, a novel activation function is here proposed for the first time. Our aim is to improve the performance of Convolutional Neural Networks in small…
Fundação Oswaldo Cruz, Instituto Gonçalo Moniz（奥斯瓦尔多·克鲁兹基金会，贡萨洛·莫尼兹研究所）；Universidade Federal de Estadual de Feira de Santana（费拉迪圣安娜州立联邦大学）
This letter presente a comment on the paper Prediction of Kidney Function from Biopsy Images using Convolutional Neural Networks by Ledbetter et al. (2017)
In this technical report, we investigate extending convolutional neural networks to the setting where functions are not sampled in a grid pattern. We show that by treating the samples as the average of a function within a cell, we can find a natural equivalent of most layers used in CNN. We also present an algorithm for running inference for these models exactly using standard convex geometry algorithms.
Quantitative analysis of brain MRI at the age of 6 months is difficult because of the limited contrast between white matter and gray matter. In this study, we use a dilated triplanar convolutional neural network in combination with a non-dilated 3D convolutional neural network for the segmentation of white matter, gray matter and cerebrospinal fluid in infant brain MR images, as provided by the MICCAI grand challenge on 6-month infant brain MRI segmentation.
Convolutional neural networks have become a popular research in the field of finger vein recognition because of their powerful image feature representation. However, most researchers focus on improving the performance of the network by increasing the CNN depth and width, which often requires high computational effort. Moreover, we can notice that not only the importance of pixels in different channels is different, but also the importance of pixels in different positions of the same channel is different. To reduce the computational effort and to take into account the different importance of pi…
Convolutional Neural Networks demonstrate high performance on ImageNet Large-Scale Visual Recognition Challenges contest. Nevertheless, the published results only show the overall performance for all image classes. There is no further analysis why certain images get worse results and how they could be improved. In this paper, we provide deep performance analysis based on different types of images and point out the weaknesses of convolutional neural networks through experiment. We design a novel multiple paths convolutional neural network, which feeds different versions of images into separated…
It is well accepted that convolutional neural networks play an important role in learning excellent features for image classification and recognition. However, in tradition they only allow adjacent layers connected, limiting integration of multi-scale information. To further improve their performance, we present a concatenating framework of shortcut convolutional neural networks. This framework can concatenate multi-scale features by shortcut connections to the fully-connected layer that is directly fed to the output layer. We do a large number of experiments to investigate performance of the …
Deep convolutional neural networks are hindered by training instability and feature redundancy towards further performance improvement. A promising solution is to impose orthogonality on convolutional filters. We develop an efficient approach to impose filter orthogonality on a convolutional layer based on the doubly block-Toeplitz matrix representation of the convolutional kernel instead of using the common kernel orthogonality approach, which we show is only necessary but not sufficient for ensuring orthogonal convolutions. Our proposed orthogonal convolution requires no additional parameter…
In image processing, it is essential to detect and track air targets, especially UAVs. In this paper, we detect the flying drone using a fisheye camera. In the field of diagnosis and classification of objects, there are always many problems that prevent the development of rapid and significant progress in this area. During the previous decades, a couple of advanced classification methods such as convolutional neural networks and support vector machines have been developed. In this study, the drone was detected using three methods of classification of convolutional neural network (CNN), support…
Image-matched nonseparable wavelets can find potential use in many applications including image classification, segmen- tation, compressive sensing, etc. This paper proposes a novel design methodology that utilizes convolutional neural net- work (CNN) to design two-channel non-separable wavelet matched to a given image. The design is proposed on quin- cunx lattice. The loss function of the convolutional neural network is setup with total squared error between the given input image to CNN and the reconstructed image at the output of CNN, leading to perfect reconstruction at the end of train- in…
George Mason University（乔治梅森大学）；Clarkson University（克拉克森大学）
Nowadays, the Convolutional Neural Networks (CNNs) have achieved impressive performance on many computer vision related tasks, such as object detection, image recognition, image retrieval, etc. These achievements benefit from the CNNs outstanding capability to learn the input features with deep layers of neuron structures and iterative training process. However, these learned features are hard to identify and interpret from a human vision perspective, causing a lack of understanding of the CNNs internal working mechanism. To improve the CNN interpretability, the CNN visualization is well utili…
Saliency detection is an important task in image processing as it can solve many problems and it usually is the first step in for other processes. Convolutional neural networks have been proved to be very effective on several image processing tasks such as classification, segmentation, semantic colorization and object manipulation. Besides, using the weights of a pretrained networks is a common practice for enhancing the accuracy of a network. In this paper a fully convolutional neural network which uses a part of VGG-16 is proposed for saliency detection in images.
Neural network compression and acceleration are widely demanded currently due to the resource constraints on most deployment targets. In this paper, through analyzing the filter activation, gradients, and visualizing the filters' functionality in convolutional neural networks, we show that the filters in higher layers learn extremely task-specific features, which are exclusive for only a small subset of the overall tasks, or even a single class. Based on such findings, we reveal the critical paths of information flow for different classes. And by their intrinsic property of exclusiveness, we p…
University of Dayton（代顿大学）；Umm Al-Qura University（乌姆·古拉大学）
This paper presents a novel method for labeling real-world neuromorphic camera sensor data by calculating the likelihood of generating an event at each pixel within a short time window, which we refer to as "event probability mask" or EPM. Its applications include (i) objective benchmarking of event denoising performance, (ii) training convolutional neural networks for noise removal called "event denoising convolutional neural network" (EDnCNN), and (iii) estimating internal neuromorphic camera parameters. We provide the first dataset (DVSNOISE20) of real-world labeled neuromorphic camera even…
Technical University of Munich（慕尼黑工业大学）；Cognition Factory GmbH（认知工厂有限责任公司）
Artificial neural networks have recently shown great results in many disciplines and a variety of applications, including natural language understanding, speech processing, games and image data generation. One particular application in which the strong performance of artificial neural networks was demonstrated is the recognition of objects in images, where deep convolutional neural networks are commonly applied. In this survey, we give a comprehensive introduction to this topic (object recognition with deep convolutional neural networks), with a strong focus on the evolution of network archite…
Model quantization is a widely used technique to compress and accelerate deep neural network (DNN) inference. Emergent DNN hardware accelerators begin to support mixed precision (1-8 bits) to further improve the computation efficiency, which raises a great challenge to find the optimal bitwidth for each layer: it requires domain experts to explore the vast design space trading off among accuracy, latency, energy, and model size, which is both time-consuming and sub-optimal. Conventional quantization algorithm ignores the different hardware architectures and quantizes all the layers in a unifor…
National Research Nuclear University "MEPhI"（国立核研究大学“MEPhI”）
We present a convolutional neural network for the classification of correlation responses obtained by correlation filters. The proposed approach can improve the accuracy of classification, as well as achieve invariance to the image classes and parameters.
School of Computer Science and Information Engineering, Hefei University of Technology（合肥工业大学计算机与信息工程学院）
The existing convolutional neural network pruning algorithms can be divided into two categories: coarse-grained clipping and fine-grained clipping. This paper proposes a coarse and fine-grained automatic pruning algorithm, which can achieve more efficient and accurate compression acceleration for convolutional neural networks. First, cluster the intermediate feature maps of the convolutional neural network to obtain the network structure after coarse-grained clipping, and then use the particle swarm optimization algorithm to iteratively search and optimize the structure. Finally, the optimal n…
Recovering an image from a noisy observation is a key problem in signal processing. Recently, it has been shown that data-driven approaches employing convolutional neural networks can outperform classical model-based techniques, because they can capture more powerful and discriminative features. However, since these methods are based on convolutional operations, they are only capable of exploiting local similarities without taking into account non-local self-similarities. In this paper we propose a convolutional neural network that employs graph-convolutional layers in order to exploit both lo…
We propose a new way of incorporating temporal information present in videos into Spatial Convolutional Neural Networks (ConvNets) trained on images, that avoids training Spatio-Temporal ConvNets from scratch. We describe several initializations of weights in 3D Convolutional Layers of Spatio-Temporal ConvNet using 2D Convolutional Weights learned from ImageNet. We show that it is important to initialize 3D Convolutional Weights judiciously in order to learn temporal representations of videos. We evaluate our methods on the UCF-101 dataset and demonstrate improvement over Spatial ConvNets.
Duke University（杜克大学）；University of Nevada, Reno（内华达大学雷诺分校）；Nanjing University（南京大学）
Neural networks have shown great performance in cognitive tasks. When deploying network models on mobile devices with limited resources, weight quantization has been widely adopted. Binary quantization obtains the highest compression but usually results in big accuracy drop. In practice, 8-bit or 16-bit quantization is often used aiming at maintaining the same accuracy as the original 32-bit precision. We observe different layers have different accuracy sensitivity of quantization. Thus judiciously selecting different precision for different layers/structures can potentially produce more effic…
Cornell University（康奈尔大学）；Augsburg University（奥格斯堡大学）；Technische Universität München（慕尼黑工业大学）；Ludwig-Maximilians-Universität München（慕尼黑大学）；Boston Children’s Hospital（波士顿儿童医院）；Harvard Medical School（哈佛医学院）
Deep convolutional neural networks are generally regarded as robust function approximators. So far, this intuition is based on perturbations to external stimuli such as the images to be classified. Here we explore the robustness of convolutional neural networks to perturbations to the internal weights and architecture of the network itself. We show that convolutional networks are surprisingly robust to a number of internal perturbations in the higher convolutional layers but the bottom convolutional layers are much more fragile. For instance, Alexnet shows less than a 30% decrease in classific…
Hunan University（湖南大学）；College of Computer Science and Electronic Engineering（计算机科学与电子工程学院）；College of Electric and Information Engineering（电气与信息工程学院）
High-resolution depth map can be inferred from a low-resolution one with the guidance of an additional high-resolution texture map of the same scene. Recently, deep neural networks with large receptive fields are shown to benefit applications such as image completion. Our insight is that super resolution is similar to image completion, where only parts of the depth values are precisely known. In this paper, we present a joint convolutional neural pyramid model with large receptive fields for joint depth map super-resolution. Our model consists of three sub-networks, two convolutional neural py…
