Most existing methods usually formulate the non-blind deconvolution problem into a maximum-a-posteriori framework and address it by manually designing a variety of regularization terms and data terms of the latent clear images. However, explicitly designing these two terms is quite challenging and usually leads to complex optimization problems which are difficult to solve. This paper proposes an effective non-blind deconvolution approach by learning discriminative shrinkage functions to model these terms implicitly. Most existing methods use deep convolutional neural networks (CNNs) or radial basis functions to learn the regularization term simply. In contrast, we formulate both the data term and regularization term and split the deconvolution model into data-related and regularization-related sub-problems according to the alternating direction method of multipliers. We explore the properties of the Maxout function and develop a deep CNN model with Maxout layers to learn discriminative shrinkage functions, which directly approximates the solutions of these two sub-problems. Moreover, the fast-Fourier-transform-based image restoration usually leads to ringing artifacts. At the same time, the conjugate-gradient-based approach is time-consuming; we develop the Conjugate Gradient Network to restore the latent clear images effectively and efficiently. Experimental results show that the proposed method performs favorably against the state-of-the-art methods in terms of efficiency and accuracy.
论文检索
输入标题、作者或关键词，从 2,893 篇学术成果中精准定位
ART-SS: An Adaptive Rejection Technique for Semi-Supervised Restoration for Adverse Weather-Affected Images
PDF ↗In recent years, convolutional neural network-based single image adverse weather removal methods have achieved significant performance improvements on many benchmark datasets. However, these methods require large amounts of clean-weather degraded image pairs for training, which is often difficult to obtain in practice. Although various weather degradation synthesis methods exist in the literature, the use of synthetically generated weather degraded images often results in sub-optimal performance on the real weatherdegraded images due to the domain gap between synthetic and real world images. To deal with this problem, various semi-supervised restoration (SSR) methods have been proposed for deraining or dehazing which learn to restore clean image using synthetically generated datasets while generalizing better using unlabeled real-world images. The performance of a semi-supervised method is essentially based on the quality of the unlabeled data. In particular, if the unlabeled data characteristics are very different from that of the labeled data, then the performance of a semi-supervised method degrades significantly. We theoretically study the effect of unlabeled data on the performance of an SSR method and develop a technique that rejects the unlabeled images that degrade the performance. Extensive experiments and ablation study show that the proposed sample rejection method increases the performance of existing SSR deraining and dehazing methods significantly. Code is available at :\textit{\href{https://github.com/rajeevyasarla/ART-SS}{https://github.com/rajeevyasarla/ART-SS}}
Conventional wisdom in blind super-resolution (SR) first estimates the unknown degradation from the low-resolution image and then exploits the degradation information for image reconstruction. Such sequential approaches suffer from two fundamental weaknesses - i.e., the lack of robustness (the performance drops when the estimated degradation is inaccurate) and the lack of transparency (network architectures are heuristic without incorporating domain knowledge). To address these issues, we propose a joint Maximum a Posterior (MAP) approach for estimating the unknown kernel and high-resolution image simultaneously. Our method first introduces uncertainty learning in the latent space when estimating the blur kernel, aiming at improving the robustness to the estimation error. Then we propose a novel SR network by unfolding the joint MAP estimator with a learned Laplacian Scale Mixture (LSM) prior and the estimated kernel. We have also developed a novel approach of estimating both the scale prior coefficient and the local means of the LSM model through a deep convolutional neural network (DCNN). All parameters of the MAP estimation algorithm and the DCNN parameters are jointly optimized through end-to-end training. Extensive experiments on both synthetic and real-world images show that our method achieves state-of-the-art performance for the task of blind image SR.
Generative adversarial networks (GANs) have proven successful in image generation tasks. However, GAN training is inherently unstable. Although many works try to stabilize it by manually modifying GAN architecture, it requires much expertise. Neural architecture search (NAS) has become an attractive solution to search GANs automatically. The early NAS-GANs search only generators to reduce search complexity but lead to a sub-optimal GAN. Some recent works try to search both generator (G) and discriminator (D), but they suffer from the instability of GAN training. To alleviate the instability, we propose an efficient two-stage evolutionary algorithm-based NAS framework to search GANs, namely EAGAN. We decouple the search of G and D into two stages, where stage-1 searches G with a fixed D and adopts the many-to-one training strategy, and stage-2 searches D with the optimal G found in stage-1 and adopts the one-to-one training and weight-resetting strategies to enhance the stability of GAN training. Both stages use the non-dominated sorting method to produce Pareto-front architectures under multiple objectives (e.g., model size, Inception Score (IS), and Fréchet Inception Distance (FID)). EAGAN is applied to the unconditional image generation task and can efficiently finish the search on the CIFAR-10 dataset in 1.2 GPU days. Our searched GANs achieve competitive results (IS=8.81±0.10, FID=9.91) on the CIFAR-10 dataset and surpass prior NAS-GANs on the STL-10 dataset (IS=10.44±0.087, FID=22.18). Source code: https://github.com/marsggbo/EAGAN.
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 then helps to conduct inference without much accuracy drop from quantization. Moreover, with a few-shot learning scenario, MEBQAT can also adapt a model to any bitwidth as well as any unseen target classes by adding conventional optimization or metric-based meta-learning. We design variants of MEBQAT to support both (1) a bitwidth-adaptive quantization scenario and (2) a new few-shot learning scenario where both quantization bitwidths and target classes are jointly adapted. We experimentally demonstrate their validity in multiple QAT schemes. By comparing their performance to (bitwidth-dedicated) QAT, existing bitwidth adaptive QAT and vanilla meta-learning, we find that merging bitwidths into meta-learning tasks achieves a higher level of robustness.
Quantization is one of the most effective methods to compress neural networks, which has achieved great success on convolutional neural networks (CNNs). Recently, vision transformers have demonstrated great potential in computer vision. However, previous post-training quantization methods performed not well on vision transformer, resulting in more than 1% accuracy drop even in 8-bit quantization. Therefore, we analyze the problems of quantization on vision transformers. We observe the distributions of activation values after softmax and GELU functions are quite different from the Gaussian distribution. We also observe that common quantization metrics, such as MSE and cosine distance, are inaccurate to determine the optimal scaling factor. In this paper, we propose the twin uniform quantization method to reduce the quantization error on these activation values. And we propose to use a Hessian guided metric to evaluate different scaling factors, which improves the accuracy of calibration with a small cost. To enable the fast quantization of vision transformers, we develop an efficient framework, PTQ4ViT. Experiments show the quantized vision transformers achieve near-lossless prediction accuracy (less than 0.5% drop at 8-bit quantization) on the ImageNet classification task.
Allocating different bit widths to different channels and quantizing them independently bring higher quantization precision and accuracy. Most of prior works use equal bit width to quantize all layers or channels, which is sub-optimal. On the other hand, it is very challenging to explore the hyperparameter space of channel bit widths, as the search space increases exponentially with the number of channels, which could be tens of thousand in a deep neural network. In this paper, we address the problem of efficiently exploring the hyperparameter space of channel bit widths. We formulate the quantization of deep neural networks as a rate-distortion optimization problem, and present an ultra-fast algorithm to search the bit allocation of channels. Our approach has only linear time complexity and can find the optimal bit allocation within a few minutes on CPU. In addition, we provide an effective way to improve the performance on target hardware platforms. We restrict the bit rate (size) of each layer to allow as many weights and activations as possible to be stored on-chip, and incorporate hardware-aware constraints into our objective function. The hardware-aware constraints do not cause additional overhead to optimization, and have very positive impact on hardware performance. Experimental results show that our approach achieves state-of-the-art results on four deep neural networks, ResNet-18, ResNet-34, ResNet-50, and MobileNet-v2, on ImageNet. Hardware simulation results demonstrate that our approach is able to bring up to 3.5x and 3.0x speedups on two deep-learning accelerators, TPU and Eyeriss, respectively.
We introduce latency-aware network acceleration (LANA)-an approach that builds on neural architecture search technique to accelerate neural networks. LANA consists of two phases: in the first phase, it trains many alternative operations for every layer of a target network using layer-wise feature map distillation. In the second phase, it solves the combinatorial selection of efficient operations using a novel constrained integer linear optimization (ILP) approach. ILP brings unique properties as it (i) performs NAS within a few seconds to minutes, (ii) easily satisfies budget constraints, (iii) works on the layer-granularity, (iv) supports a huge search space O(10^100), surpassing prior search approaches in efficacy and efficiency. In extensive experiments, we show that LANA yields efficient and accurate models constrained by a target latency budget, while being significantly faster than other techniques. We analyze three popular network architectures: EfficientNetV1, EfficientNetV2 and ResNeST, and achieve accuracy improvement (up to 3.0%) for all models when compressing larger models. LANA achieves significant speed-ups (up to 5x) with minor to no accuracy drop on GPU and CPU.
Modern deep neural networks are equipped with normalization layers such as batch normalization or layer normalization to enhance and stabilize training dynamics. If a network contains such normalization layers, the optimization objective is invariant to the scale of the neural network parameters. The scale-invariance induces the neural network’s output to be only affected by the weights’ direction and not the weights’ scale. We first find a common feature of good hyperparameter combinations on such a scale-invariant network, including learning rate, weight decay, number of data samples, and batch size. Then we observe that hyperparameter setups that lead to good performance show similar degrees of angular update during one epoch. Using a stochastic differential equation, we analyze the angular update and show how each hyperparameter affects it. With this relationship, we can derive a simple hyperparameter tuning method and apply it to the efficient hyperparameter search.
In this paper, we propose Branch-wise Activation-clipping Search Quantization (BASQ), which is a novel quantization method for low-bit activation. BASQ optimizes clip value in continuous search space while simultaneously searching L2 decay weight factor for updating clip value in discrete search space. We also propose a novel block structure for low precision that works properly on both MobileNet and ResNet structures with branch-wise searching. We evaluate the proposed methods by quantizing both weights and activations to 4-bit or lower. Contrary to the existing methods which are effective only for redundant networks, e.g., ResNet-18, or highly optimized networks, e.g., MobileNet-v2, our proposed method offers constant competitiveness on both types of networks across low precisions from 2 to 4-bits. Specifically, our 2-bit MobileNet-v2 offers top-1 accuracy of 64.71% on ImageNet, outperforming the existing method by a large margin (2.8%), and our 4-bit MobileNet-v2 gives 71.98% which is comparable to the full-precision accuracy 71.88% while our uniform quantization method offers comparable accuracy of 2-bit ResNet-18 to the state-of-the-art non-uniform quantization method.
Explicit Model Size Control and Relaxation via Smooth Regularization for Mixed-Precision Quantization
PDF ↗While Deep Neural Networks (DNNs) quantization leads to a significant reduction in computational and storage costs, it reduces model capacity and therefore, usually leads to an accuracy drop. One of the possible ways to overcome this issue is to use different quantization bit-widths for different layers. The main challenge of the mixed-precision approach is to define the bit-widths for each layer, while staying under memory and latency requirements. Motivated by this challenge, we introduce a novel technique for explicit complexity control of DNNs quantized to mixed-precision, which uses smooth optimization on the surface containing neural networks of constant size. Furthermore, we introduce a family of smooth quantization regularizers, which can be used jointly with our complexity control method for both post-training mixed-precision quantization and quantization-aware training. Our approach can be applied to any neural network architecture. Experiments show that the proposed techniques reach state-of-the-art results.
Neural network quantization techniques have been a prevailing way to reduce the inference time and storage cost of full-precision models for mobile devices. However, they still suffer from accuracy degradation due to inappropriate gradients in the optimization phase, especially for low-bit precision network and low-level vision tasks. To alleviate this issue, this paper defines a family of equivalent smooth regularizers for neural network quantization, named as SQR, which represents the equivalent of actual quantization error. Based on the definition, we propose a novel QSin regularizer as an instance to evaluate the performance of SQR, and also build up an algorithm to train the network for integer weight and activation. Extensive experimental results on classification and SR tasks reveal that the proposed method achieves higher accuracy than other prominent quantization approaches. Especially for SR task, our method alleviates the plaid artifacts effectively for quantized networks in terms of visual quality.
Towards Ultra Low Latency Spiking Neural Networks for Vision and Sequential Tasks Using Temporal Pruning
PDF ↗Spiking Neural Networks (SNNs) can be energy efficient alternatives to commonly used deep neural networks (DNNs). However, computation over multiple timesteps increases latency and energy and incurs memory access overhead of membrane potentials. Hence, latency reduction is pivotal to obtain SNNs with high energy efficiency. But, reducing latency can have an adverse effect on accuracy. To optimize the accuracy-energy-latency trade-off, we propose a temporal pruning method which starts with an SNN of T timesteps, and reduces T every iteration of training, with threshold and leak as trainable parameters. This results in a continuum of SNNs from T timesteps, all the way up to unit timestep. Training SNNs directly with 1 timestep results in convergence failure due to layerwise spike vanishing and difficulty in finding optimum thresholds. The proposed temporal pruning overcomes this by enabling the learning of suitable layerwise thresholds with backpropagation by maintaining sufficient spiking activity. Using the proposed algorithm, we achieve top-1 accuracy of 93.05%, 70.15% and 69.00% on CIFAR-10, CIFAR-100 and ImageNet, respectively with VGG16, in just 1 timestep. Note, SNNs with leaky-integrate-and-fire (LIF) neurons behave as Recurrent Neural Networks (RNNs), with the membrane potential retaining information of previous inputs. The proposed SNNs also enable performing sequential tasks such as reinforcement learning on Cartpole and Atari pong environments using only 1 to 5 timesteps.
Quantization is a very effective optimization technique to reduce hardware cost and memory footprint of deep neural network (DNN) accelerators. In particular, post-training quantization (PTQ) is often preferred as it does not require a full dataset or costly retraining. However, performance of PTQ lags significantly behind that of quantization-aware training especially for low-precision networks (<= 4-bit). In this paper we propose a novel PTQ scheme to bridge the gap, with minimal impact on hardware cost. The main idea of our scheme is to increase arithmetic precision while retaining the same representational precision. The excess arithmetic precision enables us to better match the input data distribution while also presenting a new optimization problem, to which we propose a novel search-based solution. Our scheme is based on logarithmic-scale quantization, which can help reduce hardware cost through the use of shifters instead of multipliers. Our evaluation results using various DNN models on challenging computer vision tasks (image classification, object detection, semantic segmentation) show superior accuracy compared with the state-of-the-art PTQ methods at various low-bit precisions.
When optimizing sequentially incoming tasks, deep neural networks generally suffer from catastrophic forgetting due to their lack of ability to maintain knowledge from old tasks. This may lead to a significant performance drop of the previously learned tasks. To alleviate this problem, studies on continual learning have been conducted as a countermeasure. Nevertheless, it suffers from an increase in computational cost due to the expansion of the network size or a change in knowledge that is favorably linked to previous tasks. In this work, we propose a novel approach to differentiate helpful and harmful information for old tasks using a model search to learn a current task effectively. Given a new task, the proposed method discovers an underlying association knowledge from old tasks, which can provide additional support in acquiring the new task knowledge. In addition, by introducing a sensitivity measure to the loss of the current task from the associated tasks, we find cooperative relations between tasks while alleviating harmful interference. We apply the proposed approach to both task- and class-incremental scenarios in continual learning, using a wide range of datasets from small to large scales. Experimental results show that the proposed method outperforms a large variety of continual learning approaches for the experiments while effectively alleviating catastrophic forgetting.
For practical deep neural network design on mobile devices, it is essential to consider the constraints incurred by the computational resources and the inference latency in various applications. Among deep network acceleration approaches, pruning is a widely adopted practice to balance the computational resource consumption and the accuracy, where unimportant connections can be removed either channel-wisely or randomly with a minimal impact on model accuracy. The coarse-grained channel pruning instantly results in a significant latency reduction, while the fine-grained weight pruning is more flexible to retain accuracy. In this paper, we present a unified framework for the Joint Channel pruning and Weight pruning, named JCW, which achieves an optimal pruning proportion between channel and weight pruning. To fully optimize the trade-off between latency and accuracy, we further develop a tailored multi-objective evolutionary algorithm in the JCW framework, which enables one single round search to obtain the optimal candidate architectures for various deployment requirements. Extensive experiments demonstrate that the JCW achieves a better trade-off between the latency and accuracy against previous state-of-the-art pruning methods on the ImageNet classification dataset.
This paper studies the Binary Neural Networks (BNNs) in which weights and activations are both binarized into 1-bit values, thus greatly reducing the memory usage and computational complexity. Since the modern deep neural networks are of sophisticated design with complex architecture for the accuracy reason, the diversity on distributions of weights and activations is very high. Therefore, the conventional sign function cannot be well used for effectively binarizing full precision values in BNNs. To this end, we present a simple yet effective approach called AdaBin to adaptively obtain the optimal binary sets {b_1, b_2} (b_1, b_2 belong to R) of weights and activations for each layer instead of a fixed set (i.e., {-1, +1}). In this way, the proposed method can better fit different distributions and increase the representation ability of binarized features. In practice, we use the center position and distance of 1-bit values to define a new binary quantization function. For the weights, we propose an equalization method to align the symmetrical center of binary distribution to real-valued distribution, and minimize the Kullback-Leibler divergence of them. Meanwhile, we introduce a gradient-based optimization method to get these two parameters for activations, which are jointly trained in an end-to-end manner. Experimental results on benchmark models and datasets demonstrate that the proposed AdaBin is able to achieve state-of-the-art performance. For instance, we obtain a 66.4% Top-1 accuracy on the ImageNet using ResNet-18 architecture, and a 69.4 mAP on PASCAL VOC using SSD300.
In this paper, we propose a novel channel pruning method for compression and acceleration of Convolutional Neural Networks (CNNs). Many existing channel pruning works try to discover compact sub-networks by optimizing a regularized loss function through differentiable operations. Usually, a learnable parameter is used to characterize each channel, which entangles the width and channel importance. In this setting, the FLOPs or parameter constraints implicitly restrict the search space of the pruned model. To solve the aforementioned problems, we propose optimizing each layer’s width by relaxing the hard equality constraint used in previous works. The relaxation is inspired by the definition of the top-$k$ operation. By doing so, we partially disentangle the learning of width and channel importance, which enables independent parametrization for width and importance and makes pruning more flexible. We also introduce soft top-$k$ to improve the learning of width. Moreover, to make pruning more efficient, we use two neural networks to parameterize the importance and width. The importance generation network considers both inter-channel and inter-layer relationships. The width generation network has similar functions. In addition, our method can be easily optimized by popular SGD methods, which enjoys the benefits of differentiable pruning. Extensive experiments on CIFAR-10 and ImageNet show that our method is competitive with state-of-the-art methods.
Self-attention based models such as vision transformers (ViTs) have emerged as a very competitive architecture alternative to convolutional neural networks (CNNs) in computer vision. Despite increasingly stronger variants with ever-higher recognition accuracies, due to the quadratic complexity of self-attention, existing ViTs are typically demanding in computation and model size. Although several successful design choices (e.g., the convolutions and hierarchical multi-stage structure) of prior CNNs have been reintroduced into recent ViTs, they are still not sufficient to meet the limited resource requirements of mobile devices. This motivates a very recent attempt to develop light ViTs based on the state-of-the-art MobileNet-v2, but still leaves a performance gap behind. In this work, pushing further along this under-studied direction we introduce EdgeViTs, a new family of light-weight ViTs that, for the first time, enable attention-based vision models to compete with the best light-weight CNNs in the tradeoff between accuracy and on-device efficiency. This is realized by introducing a highly cost-effective local-global-local (LGL) information exchange bottleneck based on optimal integration of self-attention and convolutions. For device-dedicated evaluation, rather than relying on inaccurate proxies like the number of FLOPs or parameters, we adopt a practical approach of focusing directly on on-device latency and, for the first time, energy efficiency. Extensive experiments on image classification, object detection, and semantic segmentation validate the high efficiency of our EdgeViTs when compared to the state-of-the-art efficient CNNs and ViTs in terms of accuracy-efficiency tradeoff on mobile hardware. Specifically, we show that our models are Pareto-optimal when both accuracy-latency and accuracy-energy tradeoffs are considered, achieving strict dominance over other ViTs in almost all cases and competing with the most efficient CNNs.
The exponentially large discrete search space in mixed-precision quantization (MPQ) makes it hard to determine the optimal bit-width for each layer. Previous works usually resort to iterative search methods on the training set, which consume hundreds or even thousands of GPU-hours. In this study, we reveal that some unique learnable parameters in quantization, namely the scale factors in the quantizer, can serve as importance indicators of a layer, reflecting the contribution of that layer to the final accuracy at certain bit-widths. These importance indicators naturally perceive the numerical transformation during quantization-aware training, which can precisely provide quantization sensitivity metrics of layers. However, a deep network always contains hundreds of such indicators, and training them one by one would lead to an excessive time cost. To overcome this issue, we propose a joint training scheme that can obtain all indicators at once. It considerably speeds up the indicators training process by parallelizing the original sequential training processes. With these learned importance indicators, we formulate the MPQ search problem as a one-time integer linear programming (ILP) problem. That avoids the iterative search and significantly reduces search time without limiting the bit-width search space. For example, MPQ search on ResNet18 with our indicators takes only 0.06 seconds, which improves time efficiency exponentially compared to iterative search methods. Also, extensive experiments show our approach can achieve SOTA accuracy on ImageNet for far-ranging models with various constraints (e.g., BitOps, compress rate).