PyTorchKR Papers

최신 논문

총 10,163편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:1909.13579공식 코드 · Papers with Code

Meta-learning algorithms for Few-Shot Computer Vision

Few-Shot Learning is the challenge of training a model with only a small amount of data. Many solutions to this problem use meta-learning algorithms, i.e. algorithms that learn to learn. By sampling few-shot tasks from a larger dataset, we can teach these algorithms to solve new, unseen tasks. This document reports my work on meta-learning algorithms for Few-Shot Computer Vision. This work was done during my internship at Sicara, a French company building image recognition solutions for businesses. It contains: 1. an extensive review of the state-of-the-art in few-shot computer vision; 2. a benchmark of meta-learning algorithms for few-shot image classification; 3. the introduction to a novel meta-learning algorithm for few-shot object detection, which is still in development.

arXiv:1909.12741

Impact of Low-bitwidth Quantization on the Adversarial Robustness for Embedded Neural Networks

As the will to deploy neural networks models on embedded systems grows, and considering the related memory footprint and energy consumption issues, finding lighter solutions to store neural networks such as weight quantization and more efficient inference methods become major research topics. Parallel to that, adversarial machine learning has risen recently with an impressive and significant attention, unveiling some critical flaws of machine learning models, especially neural networks. In particular, perturbed inputs called adversarial examples have been shown to fool a model into making incorrect predictions. In this article, we investigate the adversarial robustness of quantized neural networks under different threat models for a classical supervised image classification task. We show that quantization does not offer any robust protection, results in severe form of gradient masking and advance some hypotheses to explain it. However, we experimentally observe poor transferability capacities which we explain by quantization value shift phenomenon and gradient misalignment and explore how these results can be exploited with an ensemble-based defense.

arXiv:1909.12699공식 코드 · Papers with Code

Urban Sound Tagging using Convolutional Neural Networks

In this paper, we propose a framework for environmental sound classification in a low-data context (less than 100 labeled examples per class). We show that using pre-trained image classification models along with the usage of data augmentation techniques results in higher performance over alternative approaches. We applied this system to the task of Urban Sound Tagging, part of the DCASE 2019. The objective was to label different sources of noise from raw audio data. A modified form of MobileNetV2, a convolutional neural network (CNN) model was trained to classify both coarse and fine tags jointly. The proposed model uses log-scaled Mel-spectrogram as the representation format for the audio data. Mixup, Random erasing, scaling, and shifting are used as data augmentation techniques. A second model that uses scaled labels was built to account for human errors in the annotations. The proposed model achieved the first rank on the leaderboard with Micro-AUPRC values of 0.751 and 0.860 on fine and coarse tags, respectively.

arXiv:1909.12473공식 코드 · Papers with Code

Noisy Batch Active Learning with Deterministic Annealing

We study the problem of training machine learning models incrementally with batches of samples annotated with noisy oracles. We select each batch of samples that are important and also diverse via clustering and importance sampling. More importantly, we incorporate model uncertainty into the sampling probability to compensate for poor estimation of the importance scores when the training data is too small to build a meaningful model. Experiments on benchmark image classification datasets (MNIST, SVHN, CIFAR10, and EMNIST) show improvement over existing active learning strategies. We introduce an extra denoising layer to deep networks to make active learning robust to label noises and show significant improvements.

arXiv:1909.12446

Interpreting Undesirable Pixels for Image Classification on Black-Box Models

In an effort to interpret black-box models, researches for developing explanation methods have proceeded in recent years. Most studies have tried to identify input pixels that are crucial to the prediction of a classifier. While this approach is meaningful to analyse the characteristic of blackbox models, it is also important to investigate pixels that interfere with the prediction. To tackle this issue, in this paper, we propose an explanation method that visualizes undesirable regions to classify an image as a target class. To be specific, we divide the concept of undesirable regions into two terms: (1) factors for a target class, which hinder that black-box models identify intrinsic characteristics of a target class and (2) factors for non-target classes that are important regions for an image to be classified as other classes. We visualize such undesirable regions on heatmaps to qualitatively validate the proposed method. Furthermore, we present an evaluation metric to provide quantitative results on ImageNet.

arXiv:1909.11799공식 코드 · Papers with Code

Manifold Oblique Random Forests: Towards Closing the Gap on Convolutional Deep Networks

Decision forests (Forests), in particular random forests and gradient boosting trees, have demonstrated state-of-the-art accuracy compared to other methods in many supervised learning scenarios. In particular, Forests dominate other methods in tabular data, that is, when the feature space is unstructured, so that the signal is invariant to a permutation of the feature indices. However, in structured data lying on a manifold (such as images, text, and speech) deep networks (Networks), specifically convolutional deep networks (ConvNets), tend to outperform Forests. We conjecture that at least part of the reason for this is that the input to Networks is not simply the feature magnitudes, but also their indices. In contrast, naive Forest implementations fail to explicitly consider feature indices. A recently proposed Forest approach demonstrates that Forests, for each node, implicitly sample a random matrix from some specific distribution. These Forests, like some classes of Networks, learn by partitioning the feature space into convex polytopes corresponding to linear functions. We build on that approach and show that one can choose distributions in a manifold-aware fashion to incorporate feature locality. We demonstrate the empirical performance on data whose features live on three different manifolds: a torus, images, and time-series. Moreover, we demonstrate its strength in multivariate simulated settings and also show superiority in predicting surgical outcome in epilepsy patients and predicting movement direction from raw stereotactic EEG data from non-motor brain regions. In all simulations and real data, Manifold Oblique Random Forest (MORF) algorithm outperforms approaches that ignore feature space structure and challenges the performance of ConvNets. Moreover, MORF runs fast and maintains interpretability and theoretical justification.

arXiv:1909.12205

Adaptive Binary-Ternary Quantization

Neural network models are resource hungry. It is difficult to deploy such deep networks on devices with limited resources, like smart wearables, cellphones, drones, and autonomous vehicles. Low bit quantization such as binary and ternary quantization is a common approach to alleviate this resource requirements. Ternary quantization provides a more flexible model and outperforms binary quantization in terms of accuracy, however doubles the memory footprint and increases the computational cost. Contrary to these approaches, mixed quantized models allow a trade-off between accuracy and memory footprint. In such models, quantization depth is often chosen manually, or is tuned using a separate optimization routine. The latter requires training a quantized network multiple times. Here, we propose an adaptive combination of binary and ternary quantization, namely Smart Quantization (SQ), in which the quantization depth is modified directly via a regularization function, so that the model is trained only once. Our experimental results show that the proposed method adapts quantization depth successfully while keeping the model accuracy high on MNIST and CIFAR10 benchmarks.

arXiv:1909.12117

Balanced Binary Neural Networks with Gated Residual

Binary neural networks have attracted numerous attention in recent years. However, mainly due to the information loss stemming from the biased binarization, how to preserve the accuracy of networks still remains a critical issue. In this paper, we attempt to maintain the information propagated in the forward process and propose a Balanced Binary Neural Networks with Gated Residual (BBG for short). First, a weight balanced binarization is introduced to maximize information entropy of binary weights, and thus the informative binary weights can capture more information contained in the activations. Second, for binary activations, a gated residual is further appended to compensate their information loss during the forward process, with a slight overhead. Both techniques can be wrapped as a generic network module that supports various network architectures for different tasks including classification and detection. We evaluate our BBG on image classification tasks over CIFAR-10/100 and ImageNet and on detection task over Pascal VOC. The experimental results show that BBG-Net performs remarkably well across various network architectures such as VGG, ResNet and SSD with the superior performance over state-of-the-art methods in terms of memory consumption, inference speed and accuracy.

arXiv:1909.12111

Two-stage Image Classification Supervised by a Single Teacher Single Student Model

The two-stage strategy has been widely used in image classification. However, these methods barely take the classification criteria of the first stage into consideration in the second prediction stage. In this paper, we propose a novel two-stage representation method (TSR), and convert it to a Single-Teacher Single-Student (STSS) problem in our two-stage image classification framework. We seek the nearest neighbours of the test sample to choose candidate target classes. Meanwhile, the first stage classifier is formulated as the teacher, which holds the classification scores. The samples of the candidate classes are utilized to learn a student classifier based on L2-minimization in the second stage. The student will be supervised by the teacher classifier, which approves the student only if it obtains a higher score. In actuality, the proposed framework generates a stronger classifier by staging two weaker classifiers in a novel way. The experiments conducted on several face and object databases show that our proposed framework is effective and outperforms multiple popular classification methods.

arXiv:1909.11953

Hyperspectral Image Classification With Context-Aware Dynamic Graph Convolutional Network

In hyperspectral image (HSI) classification, spatial context has demonstrated its significance in achieving promising performance. However, conventional spatial context-based methods simply assume that spatially neighboring pixels should correspond to the same land-cover class, so they often fail to correctly discover the contextual relations among pixels in complex situations, and thus leading to imperfect classification results on some irregular or inhomogeneous regions such as class boundaries. To address this deficiency, we develop a new HSI classification method based on the recently proposed Graph Convolutional Network (GCN), as it can flexibly encode the relations among arbitrarily structured non-Euclidean data. Different from traditional GCN, there are two novel strategies adopted by our method to further exploit the contextual relations for accurate HSI classification. First, since the receptive field of traditional GCN is often limited to fairly small neighborhood, we proposed to capture long range contextual relations in HSI by performing successive graph convolutions on a learned region-induced graph which is transformed from the original 2D image grids. Second, we refine the graph edge weight and the connective relationships among image regions by learning the improved adjacency matrix and the 'edge filter', so that the graph can be gradually refined to adapt to the representations generated by each graph convolutional layer. Such updated graph will in turn result in accurate region representations, and vice versa. The experiments carried out on three real-world benchmark datasets demonstrate that the proposed method yields significant improvement in the classification performance when compared with some state-of-the-art approaches.

arXiv:1909.11519공식 코드 · Papers with Code

Gated Channel Transformation for Visual Recognition

In this work, we propose a generally applicable transformation unit for visual recognition with deep convolutional neural networks. This transformation explicitly models channel relationships with explainable control variables. These variables determine the neuron behaviors of competition or cooperation, and they are jointly optimized with the convolutional weight towards more accurate recognition. In Squeeze-and-Excitation (SE) Networks, the channel relationships are implicitly learned by fully connected layers, and the SE block is integrated at the block-level. We instead introduce a channel normalization layer to reduce the number of parameters and computational complexity. This lightweight layer incorporates a simple l2 normalization, enabling our transformation unit applicable to operator-level without much increase of additional parameters. Extensive experiments demonstrate the effectiveness of our unit with clear margins on many vision tasks, i.e., image classification on ImageNet, object detection and instance segmentation on COCO, video classification on Kinetics.

arXiv:1909.11498

Non-imaging single-pixel sensing with optimized binary modulation

The conventional high-level sensing techniques require high-fidelity images as input to extract target features, which are produced by either complex imaging hardware or high-complexity reconstruction algorithms. In this letter, we propose single-pixel sensing (SPS) that performs high-level sensing directly from coupled measurements of a single-pixel detector, without the conventional image acquisition and reconstruction process. The technique consists of three steps including binary light modulation that can be physically implemented at $\sim$22kHz, single-pixel coupled detection owning wide working spectrum and high signal-to-noise ratio, and end-to-end deep-learning based sensing that reduces both hardware and software complexity. Besides, the binary modulation is trained and optimized together with the sensing network, which ensures least required measurements and optimal sensing accuracy. The effectiveness of SPS is demonstrated on the classification task of handwritten MNIST dataset, and 96.68% classification accuracy at $\sim$1kHz is achieved. The reported single-pixel sensing technique is a novel framework for highly efficient machine intelligence.

arXiv:1909.11380

Beyond image classification: zooplankton identification with deep vector space embeddings

Zooplankton images, like many other real world data types, have intrinsic properties that make the design of effective classification systems difficult. For instance, the number of classes encountered in practical settings is potentially very large, and classes can be ambiguous or overlap. In addition, the choice of taxonomy often differs between researchers and between institutions. Although high accuracy has been achieved in benchmarks using standard classifier architectures, biases caused by an inflexible classification scheme can have profound effects when the output is used in ecosystem assessments and monitoring. Here, we propose using a deep convolutional network to construct a vector embedding of zooplankton images. The system maps (embeds) each image into a high-dimensional Euclidean space so that distances between vectors reflect semantic relationships between images. We show that the embedding can be used to derive classifications with comparable accuracy to a specific classifier, but that it simultaneously reveals important structures in the data. Furthermore, we apply the embedding to new classes previously unseen by the system, and evaluate its classification performance in such cases. Traditional neural network classifiers perform well when the classes are clearly defined a priori and have sufficiently large labeled data sets available. For practical cases in ecology as well as in many other fields this is not the case, and we argue that the vector embedding method presented here is a more appropriate approach.

arXiv:1909.11212

Augmenting the Pathology Lab: An Intelligent Whole Slide Image Classification System for the Real World

Standard of care diagnostic procedure for suspected skin cancer is microscopic examination of hematoxylin \& eosin stained tissue by a pathologist. Areas of high inter-pathologist discordance and rising biopsy rates necessitate higher efficiency and diagnostic reproducibility. We present and validate a deep learning system which classifies digitized dermatopathology slides into 4 categories. The system is developed using 5,070 images from a single lab, and tested on an uncurated set of 13,537 images from 3 test labs, using whole slide scanners manufactured by 3 different vendors. The system's use of deep-learning-based confidence scoring as a criterion to consider the result as accurate yields an accuracy of up to 98\%, and makes it adoptable in a real-world setting. Without confidence scoring, the system achieved an accuracy of 78\%. We anticipate that our deep learning system will serve as a foundation enabling faster diagnosis of skin cancer, identification of cases for specialist review, and targeted diagnostic classifications.

arXiv:1909.11193

Scaling-Translation-Equivariant Networks with Decomposed Convolutional Filters

Encoding the scale information explicitly into the representation learned by a convolutional neural network (CNN) is beneficial for many computer vision tasks especially when dealing with multiscale inputs. We study, in this paper, a scaling-translation-equivariant (ST-equivariant) CNN with joint convolutions across the space and the scaling group, which is shown to be both sufficient and necessary to achieve equivariance for the regular representation of the scaling-translation group ST . To reduce the model complexity and computational burden, we decompose the convolutional filters under two pre-fixed separable bases and truncate the expansion to low-frequency components. A further benefit of the truncated filter expansion is the improved deformation robustness of the equivariant representation, a property which is theoretically analyzed and empirically verified. Numerical experiments demonstrate that the proposed scaling-translation-equivariant network with decomposed convolutional filters (ScDCFNet) achieves significantly improved performance in multiscale image classification and better interpretability than regular CNNs at a reduced model size.

arXiv:1909.11155

Anchor Loss: Modulating Loss Scale based on Prediction Difficulty

We propose a novel loss function that dynamically rescales the cross entropy based on prediction difficulty regarding a sample. Deep neural network architectures in image classification tasks struggle to disambiguate visually similar objects. Likewise, in human pose estimation symmetric body parts often confuse the network with assigning indiscriminative scores to them. This is due to the output prediction, in which only the highest confidence label is selected without taking into consideration a measure of uncertainty. In this work, we define the prediction difficulty as a relative property coming from the confidence score gap between positive and negative labels. More precisely, the proposed loss function penalizes the network to avoid the score of a false prediction being significant. To demonstrate the efficacy of our loss function, we evaluate it on two different domains: image classification and human pose estimation. We find improvements in both applications by achieving higher accuracy compared to the baseline methods.

arXiv:1909.10818

Constrained deep neural network architecture search for IoT devices accounting hardware calibration

Deep neural networks achieve outstanding results in challenging image classification tasks. However, the design of network topologies is a complex task and the research community makes a constant effort in discovering top-accuracy topologies, either manually or employing expensive architecture searches. In this work, we propose a unique narrow-space architecture search that focuses on delivering low-cost and fast executing networks that respect strict memory and time requirements typical of Internet-of-Things (IoT) near-sensor computing platforms. Our approach provides solutions with classification latencies below 10ms running on a $35 device with 1GB RAM and 5.6GFLOPS peak performance. The narrow-space search of floating-point models improves the accuracy on CIFAR10 of an established IoT model from 70.64% to 74.87% respecting the same memory constraints. We further improve the accuracy to 82.07% by including 16-bit half types and we obtain the best accuracy of 83.45% by extending the search with model optimized IEEE 754 reduced types. To the best of our knowledge, we are the first that empirically demonstrate on over 3000 trained models that running with reduced precision pushes the Pareto optimal front by a wide margin. Under a given memory constraint, accuracy is improved by over 7% points for half and over 1% points further for running with the best model individual format.

arXiv:1909.10811

Image Recognition using Region Creep

This paper describes a new type of auto-associative image classifier that uses a shallow architecture with a very quick learning phase. The image is parsed into smaller areas and each area is saved directly for a region, along with the related output category. When a new image is presented, a direct match with each region is made and the best matching areas returned. Each area stores a list of the categories it belongs to, where there is a one-to-many relation between the input region and the output categories. The image classification process sums the category lists to return a preferred category for the whole image. These areas can overlap with each other and when moving from a region to its neighbours, there is likely to be only small changes in the area image part. It would therefore be possible to guess what the best image area is for one region by cumulating the results of its neighbours. This associative feature is being called 'Region Creep' and the cumulated region can be compared with train cases instead, when a suitable match is not found. Rules can be included and state that: if one set of pixels are present, another set should either be removed or should also be present, where this is across the whole image. The memory problems with a traditional auto-associative network may be less with this version and tests on a set of hand-written numbers have produced state-of-the-art results.

arXiv:1909.10783공식 코드 · Papers with Code

PolSAR Image Classification Based on Dilated Convolution and Pixel-Refining Parallel Mapping network in the Complex Domain

Efficient and accurate polarimetric synthetic aperture radar (PolSAR) image classification with a limited number of prior labels is always full of challenges. For general supervised deep learning classification algorithms, the pixel-by-pixel algorithm achieves precise yet inefficient classification with a small number of labeled pixels, whereas the pixel mapping algorithm achieves efficient yet edge-rough classification with more prior labels required. To take efficiency, accuracy and prior labels into account, we propose a novel pixel-refining parallel mapping network in the complex domain named CRPM-Net and the corresponding training algorithm for PolSAR image classification. CRPM-Net consists of two parallel sub-networks: a) A transfer dilated convolution mapping network in the complex domain (C-Dilated CNN) activated by a complex cross-convolution neural network (Cs-CNN), which is aiming at precise localization, high efficiency and the full use of phase information; b) A complex domain encoder-decoder network connected parallelly with C-Dilated CNN, which is to extract more contextual semantic features. Finally, we design a two-step algorithm to train the Cs-CNN and CRPM-Net with a small number of labeled pixels for higher accuracy by refining misclassified labeled pixels. We verify the proposed method on AIRSAR and E-SAR datasets. The experimental results demonstrate that CRPM-Net achieves the best classification results and substantially outperforms some latest state-of-the-art approaches in both efficiency and accuracy for PolSAR image classification. The source code and trained models for CRPM-Net is available at: https://github.com/PROoshio/CRPM-Net.

arXiv:1909.10708

Unsupervised Deep Features for Privacy Image Classification

Sharing images online poses security threats to a wide range of users due to the unawareness of privacy information. Deep features have been demonstrated to be a powerful representation for images. However, deep features usually suffer from the issues of a large size and requiring a huge amount of data for fine-tuning. In contrast to normal images (e.g., scene images), privacy images are often limited because of sensitive information. In this paper, we propose a novel approach that can work on limited data and generate deep features of smaller size. For training images, we first extract the initial deep features from the pre-trained model and then employ the K-means clustering algorithm to learn the centroids of these initial deep features. We use the learned centroids from training features to extract the final features for each testing image and encode our final features with the triangle encoding. To improve the discriminability of the features, we further perform the fusion of two proposed unsupervised deep features obtained from different layers. Experimental results show that the proposed features outperform state-of-the-art deep features, in terms of both classification accuracy and testing time.