1.Camera Model Anonymisation with Augmented cGANs ⬇️
The model of camera that was used to capture a particular photographic image (model attribution) can be inferred from model-specific artefacts present within the image. Typically these artefacts are found in high-frequency pixel patterns, rather than image content. Model anonymisation is the process of transforming these artefacts such that the apparent capture model is changed. Improved methods for attribution and anonymisation are important for improving digital forensics, and understanding its limits. Through conditional adversarial training, we present an approach for learning these transformations. Significantly, we augment the objective with the losses from pre-trained auxiliary model attribution classifiers that constrain the generator to not only synthesise discriminative high-frequency artefacts, but also salient image-based artefacts lost during image content suppression. Quantitative comparisons against a recent representative approach demonstrate the efficacy of our framework in a non-interactive black-box setting.
2.MAST: A Memory-Augmented Self-supervised Tracker ⬇️
Recent interest in self-supervised dense tracking has yielded rapid progress, but performance still remains far from supervised methods. We propose a dense tracking model trained on videos without any annotations that surpasses previous self-supervised methods on existing benchmarks by a significant margin (+15%), and achieves performance comparable to supervised methods. In this paper, we first reassess the traditional choices used for self-supervised training and reconstruction loss by conducting thorough experiments that finally elucidate the optimal choices. Second, we further improve on existing methods by augmenting our architecture with a crucial memory component. Third, we benchmark on large-scale semi-supervised video object segmentation(aka. dense tracking), and propose a new metric: generalizability. Our first two contributions yield a self-supervised network that for the first time is competitive with supervised methods on standard evaluation metrics of dense tracking. When measuring generalizability, we show self-supervised approaches are actually superior to the majority of supervised methods. We believe this new generalizability metric can better capture the real-world use-cases for dense tracking, and will spur new interest in this research direction.
3.Computational optimization of convolutional neural networks using separated filters architecture ⬇️
This paper considers a convolutional neural network transformation that reduces computation complexity and thus speedups neural network processing. Usage of convolutional neural networks (CNN) is the standard approach to image recognition despite the fact they can be too computationally demanding, for example for recognition on mobile platforms or in embedded systems. In this paper we propose CNN structure transformation which expresses 2D convolution filters as a linear combination of separable filters. It allows to obtain separated convolutional filters by standard training algorithms. We study the computation efficiency of this structure transformation and suggest fast implementation easily handled by CPU or GPU. We demonstrate that CNNs designed for letter and digit recognition of proposed structure show 15% speedup without accuracy loss in industrial image recognition system. In conclusion, we discuss the question of possible accuracy decrease and the application of proposed transformation to different recognition problems. convolutional neural networks, computational optimization, separable filters, complexity reduction.
4.Towards Bounding-Box Free Panoptic Segmentation ⬇️
In this work we introduce a new bounding-box free network (BBFNet) for panoptic segmentation. Panoptic segmentation is an ideal problem for a bounding-box free approach as it already requires per-pixel semantic class labels. We use this observation to exploit class boundaries from an off-the-shelf semantic segmentation network and refine them to predict instance labels. Towards this goal BBFNet predicts coarse watershed levels and use it to detect large instance candidates where boundaries are well defined. For smaller instances, whose boundaries are less reliable, BBFNet also predicts instance centers by means of Hough voting followed by mean-shift to reliably detect small objects. A novel triplet loss network helps merging fragmented instances while refining boundary pixels. Our approach is distinct from previous works in panoptic segmentation that rely on a combination of a semantic segmentation network with a computationally costly instance segmentation network based on bounding boxes, such as Mask R-CNN, to guide the prediction of instance labels using a Mixture-of-Expert (MoE) approach. We benchmark our non-MoE method on Cityscapes and Microsoft COCO datasets and show competitive performance with other MoE based approaches while outperfroming exisiting non-proposal based approaches. We achieve this while been computationally more efficient in terms of number of parameters and FLOPs. Video results are provided here this https URL.
5.Voxel-Based Indoor Reconstruction From HoloLens Triangle Meshes ⬇️
Current mobile augmented reality devices are often equipped with range sensors. The Microsoft HoloLens for instance is equipped with a Time-Of-Flight (ToF) range camera providing coarse triangle meshes that can be used in custom applications. We suggest to use the triangle meshes for the automatic generation of indoor models that can serve as basis for augmenting their physical counterpart with location-dependent information. In this paper, we present a novel voxel-based approach for automated indoor reconstruction from unstructured three-dimensional geometries like triangle meshes. After an initial voxelization of the input data, rooms are detected in the resulting voxel grid by segmenting connected voxel components of ceiling candidates and extruding them downwards to find floor candidates. Semantic class labels like 'Wall', 'Wall Opening', 'Interior Object' and 'Empty Interior' are then assigned to the room voxels in-between ceiling and floor by a rule-based voxel sweep algorithm. Finally, the geometry of the detected walls and their openings is refined in voxel representation. The proposed approach is not restricted to Manhattan World scenarios and does not rely on room surfaces being planar.
6.FeatureNMS: Non-Maximum Suppression by Learning Feature Embeddings ⬇️
Most state of the art object detectors output multiple detections per object. The duplicates are removed in a post-processing step called Non-Maximum Suppression. Classical Non-Maximum Suppression has shortcomings in scenes that contain objects with high overlap: The idea of this heuristic is that a high bounding box overlap corresponds to a high probability of having a duplicate. We propose FeatureNMS to solve this problem. FeatureNMS recognizes duplicates not only based on the intersection over union between bounding boxes, but also based on the difference of feature vectors. These feature vectors can encode more information like visual appearance. Our approach outperforms classical NMS and derived approaches and achieves state of the art performance.
7.Neural arbitrary style transfer for portrait images using the attention mechanism ⬇️
Arbitrary style transfer is the task of synthesis of an image that has never been seen before, using two given images: content image and style image. The content image forms the structure, the basic geometric lines and shapes of the resulting image, while the style image sets the color and texture of the result. The word "arbitrary" in this context means the absence of any one pre-learned style. So, for example, convolutional neural networks capable of transferring a new style only after training or retraining on a new amount of data are not con-sidered to solve such a problem, while networks based on the attention mech-anism that are capable of performing such a transformation without retraining - yes. An original image can be, for example, a photograph, and a style image can be a painting of a famous artist. The resulting image in this case will be the scene depicted in the original photograph, made in the stylie of this picture. Recent arbitrary style transfer algorithms make it possible to achieve good re-sults in this task, however, in processing portrait images of people, the result of such algorithms is either unacceptable due to excessive distortion of facial features, or weakly expressed, not bearing the characteristic features of a style image. In this paper, we consider an approach to solving this problem using the combined architecture of deep neural networks with a attention mechanism that transfers style based on the contents of a particular image segment: with a clear predominance of style over the form for the background part of the im-age, and with the prevalence of content over the form in the image part con-taining directly the image of a person.
8.An interpretable classifier for high-resolution breast cancer screening images utilizing weakly supervised localization ⬇️
Medical images differ from natural images in significantly higher resolutions and smaller regions of interest. Because of these differences, neural network architectures that work well for natural images might not be applicable to medical image analysis. In this work, we extend the globally-aware multiple instance classifier, a framework we proposed to address these unique properties of medical images. This model first uses a low-capacity, yet memory-efficient, network on the whole image to identify the most informative regions. It then applies another higher-capacity network to collect details from chosen regions. Finally, it employs a fusion module that aggregates global and local information to make a final prediction. While existing methods often require lesion segmentation during training, our model is trained with only image-level labels and can generate pixel-level saliency maps indicating possible malignant findings. We apply the model to screening mammography interpretation: predicting the presence or absence of benign and malignant lesions. On the NYU Breast Cancer Screening Dataset, consisting of more than one million images, our model achieves an AUC of 0.93 in classifying breasts with malignant findings, outperforming ResNet-34 and Faster R-CNN. Compared to ResNet-34, our model is 4.1x faster for inference while using 78.4% less GPU memory. Furthermore, we demonstrate, in a reader study, that our model surpasses radiologist-level AUC by a margin of 0.11. The proposed model is available online: this https URL.
9.Few-Shot Few-Shot Learning and the role of Spatial Attention ⬇️
Few-shot learning is often motivated by the ability of humans to learn new tasks from few examples. However, standard few-shot classification benchmarks assume that the representation is learned on a limited amount of base class data, ignoring the amount of prior knowledge that a human may have accumulated before learning new tasks. At the same time, even if a powerful representation is available, it may happen in some domain that base class data are limited or non-existent. This motivates us to study a problem where the representation is obtained from a classifier pre-trained on a large-scale dataset of a different domain, assuming no access to its training process, while the base class data are limited to few examples per class and their role is to adapt the representation to the domain at hand rather than learn from scratch. We adapt the representation in two stages, namely on the few base class data if available and on the even fewer data of new tasks. In doing so, we obtain from the pre-trained classifier a spatial attention map that allows focusing on objects and suppressing background clutter. This is important in the new problem, because when base class data are few, the network cannot learn where to focus implicitly. We also show that a pre-trained network may be easily adapted to novel classes, without meta-learning.
10.NoiseBreaker: Gradual Image Denoising Guided by Noise Analysis ⬇️
Fully supervised deep-learning based denoisers are currently the most performing image denoising solutions. However, they require clean reference images. When the target noise is complex, e.g. composed of an unknown mixture of primary noises with unknown intensity, fully supervised solutions are limited by the difficulty to build a suited training set for the problem. This paper proposes a gradual denoising strategy that iteratively detects the dominating noise in an image, and removes it using a tailored denoiser. The method is shown to keep up with state of the art blind denoisers on mixture noises. Moreover, noise analysis is demonstrated to guide denoisers efficiently not only on noise type, but also on noise intensity. The method provides an insight on the nature of the encountered noise, and it makes it possible to extend an existing denoiser with new noise nature. This feature makes the method adaptive to varied denoising cases.
11.Motion Deblurring using Spatiotemporal Phase Aperture Coding ⬇️
Motion blur is a known issue in photography, as it limits the exposure time while capturing moving objects. Extensive research has been carried to compensate for it. In this work, a computational imaging approach for motion deblurring is proposed and demonstrated. Using dynamic phase-coding in the lens aperture during the image acquisition, the trajectory of the motion is encoded in an intermediate optical image. This encoding embeds both the motion direction and extent by coloring the spatial blur of each object. The color cues serve as prior information for a blind deblurring process, implemented using a convolutional neural network (CNN) trained to utilize such coding for image restoration. We demonstrate the advantage of the proposed approach over blind-deblurring with no coding and other solutions that use coded acquisition, both in simulation and real-world experiments.
12.Knowledge Integration Networks for Action Recognition ⬇️
In this work, we propose Knowledge Integration Networks (referred as KINet) for video action recognition. KINet is capable of aggregating meaningful context features which are of great importance to identifying an action, such as human information and scene context. We design a three-branch architecture consisting of a main branch for action recognition, and two auxiliary branches for human parsing and scene recognition which allow the model to encode the knowledge of human and scene for action recognition. We explore two pre-trained models as teacher networks to distill the knowledge of human and scene for training the auxiliary tasks of KINet. Furthermore, we propose a two-level knowledge encoding mechanism which contains a Cross Branch Integration (CBI) module for encoding the auxiliary knowledge into medium-level convolutional features, and an Action Knowledge Graph (AKG) for effectively fusing high-level context information. This results in an end-to-end trainable framework where the three tasks can be trained collaboratively, allowing the model to compute strong context knowledge efficiently. The proposed KINet achieves the state-of-the-art performance on a large-scale action recognition benchmark Kinetics-400, with a top-1 accuracy of 77.8%. We further demonstrate that our KINet has strong capability by transferring the Kinetics-trained model to UCF-101, where it obtains 97.8% top-1 accuracy.
13.Automated Cardiothoracic Ratio Calculation and Cardiomegaly Detection using Deep Learning Approach ⬇️
We propose an algorithm for calculating the cardiothoracic ratio (CTR) from chest X-ray films. Our approach applies a deep learning model based on U-Net with VGG16 encoder to extract lung and heart masks from chest X-ray images and calculate CTR from the extents of obtained masks. Human radiologists evaluated our CTR measurements, and
$76.5%$ were accepted to be included in medical reports without any need for adjustment. This result translates to a large amount of time and labor saved for radiologists using our automated tools.
14.Registration of multi-view point sets under the perspective of expectation-maximization ⬇️
Registration of multi-view point sets is a prerequisite for 3D model reconstruction. To solve this problem, most of previous approaches either partially explore available information or blindly utilize unnecessary information to align each point set, which may lead to the undesired results or introduce extra computation complexity. To this end, this paper consider the multi-view registration problem as a maximum likelihood estimation problem and proposes a novel multi-view registration approach under the perspective of Expectation-Maximization (EM). The basic idea of our approach is that different data points are generated by the same number of Gaussian mixture models (GMMs). For each data point in one well-aligned point set, its nearest neighbors can be searched from other well-aligned point sets to explore more available information. Then, we can suppose this data point is generated by the special GMM, which is composed of each of its nearest neighbor adhered with one Gaussian distribution. Based on this assumption, it is reasonable to define the likelihood function, which contains all rigid transformations required to be estimated for multi-view registration. Subsequently, the EM algorithm is utilized to maximize the likelihood function so as to estimate all rigid transformations. Finally, the proposed approach is tested on several bench mark data sets and compared with some state-of-the-art algorithms. Experimental results illustrate its super performance on accuracy and efficiency for the registration of multi-view point sets.
15.V4D:4D Convolutional Neural Networks for Video-level Representation Learning ⬇️
Most existing 3D CNNs for video representation learning are clip-based methods, and thus do not consider video-level temporal evolution of spatio-temporal features. In this paper, we propose Video-level 4D Convolutional Neural Networks, referred as V4D, to model the evolution of long-range spatio-temporal representation with 4D convolutions, and at the same time, to preserve strong 3D spatio-temporal representation with residual connections. Specifically, we design a new 4D residual block able to capture inter-clip interactions, which could enhance the representation power of the original clip-level 3D CNNs. The 4D residual blocks can be easily integrated into the existing 3D CNNs to perform long-range modeling hierarchically. We further introduce the training and inference methods for the proposed V4D. Extensive experiments are conducted on three video recognition benchmarks, where V4D achieves excellent results, surpassing recent 3D CNNs by a large margin.
16.EHSOD: CAM-Guided End-to-end Hybrid-Supervised Object Detection with Cascade Refinement ⬇️
Object detectors trained on fully-annotated data currently yield state of the art performance but require expensive manual annotations. On the other hand, weakly-supervised detectors have much lower performance and cannot be used reliably in a realistic setting. In this paper, we study the hybrid-supervised object detection problem, aiming to train a high quality detector with only a limited amount of fullyannotated data and fully exploiting cheap data with imagelevel labels. State of the art methods typically propose an iterative approach, alternating between generating pseudo-labels and updating a detector. This paradigm requires careful manual hyper-parameter tuning for mining good pseudo labels at each round and is quite time-consuming. To address these issues, we present EHSOD, an end-to-end hybrid-supervised object detection system which can be trained in one shot on both fully and weakly-annotated data. Specifically, based on a two-stage detector, we proposed two modules to fully utilize the information from both kinds of labels: 1) CAMRPN module aims at finding foreground proposals guided by a class activation heat-map; 2) hybrid-supervised cascade module further refines the bounding-box position and classification with the help of an auxiliary head compatible with image-level data. Extensive experiments demonstrate the effectiveness of the proposed method and it achieves comparable results on multiple object detection benchmarks with only 30% fully-annotated data, e.g. 37.5% mAP on COCO. We will release the code and the trained models.
17.Universal-RCNN: Universal Object Detector via Transferable Graph R-CNN ⬇️
The dominant object detection approaches treat each dataset separately and fit towards a specific domain, which cannot adapt to other domains without extensive retraining. In this paper, we address the problem of designing a universal object detection model that exploits diverse category granularity from multiple domains and predict all kinds of categories in one system. Existing works treat this problem by integrating multiple detection branches upon one shared backbone network. However, this paradigm overlooks the crucial semantic correlations between multiple domains, such as categories hierarchy, visual similarity, and linguistic relationship. To address these drawbacks, we present a novel universal object detector called Universal-RCNN that incorporates graph transfer learning for propagating relevant semantic information across multiple datasets to reach semantic coherency. Specifically, we first generate a global semantic pool by integrating all high-level semantic representation of all the categories. Then an Intra-Domain Reasoning Module learns and propagates the sparse graph representation within one dataset guided by a spatial-aware GCN. Finally, an InterDomain Transfer Module is proposed to exploit diverse transfer dependencies across all domains and enhance the regional feature representation by attending and transferring semantic contexts globally. Extensive experiments demonstrate that the proposed method significantly outperforms multiple-branch models and achieves the state-of-the-art results on multiple object detection benchmarks (mAP: 49.1% on COCO).
18.DivideMix: Learning with Noisy Labels as Semi-supervised Learning ⬇️
Deep neural networks are known to be annotation-hungry. Numerous efforts have been devoted to reducing the annotation cost when learning with deep networks. Two prominent directions include learning with noisy labels and semi-supervised learning by exploiting unlabeled data. In this work, we propose DivideMix, a novel framework for learning with noisy labels by leveraging semi-supervised learning techniques. In particular, DivideMix models the per-sample loss distribution with a mixture model to dynamically divide the training data into a labeled set with clean samples and an unlabeled set with noisy samples, and trains the model on both the labeled and unlabeled data in a semi-supervised manner. To avoid confirmation bias, we simultaneously train two diverged networks where each network uses the dataset division from the other network. During the semi-supervised training phase, we improve the MixMatch strategy by performing label co-refinement and label co-guessing on labeled and unlabeled samples, respectively. Experiments on multiple benchmark datasets demonstrate substantial improvements over state-of-the-art methods. Code is available at this https URL .
19.High-Order Paired-ASPP Networks for Semantic Segmenation ⬇️
Current semantic segmentation models only exploit first-order statistics, while rarely exploring high-order statistics. However, common first-order statistics are insufficient to support a solid unanimous representation. In this paper, we propose High-Order Paired-ASPP Network to exploit high-order statistics from various feature levels. The network first introduces a High-Order Representation module to extract the contextual high-order information from all stages of the backbone. They can provide more semantic clues and discriminative information than the first-order ones. Besides, a Paired-ASPP module is proposed to embed high-order statistics of the early stages into the last stage. It can further preserve the boundary-related and spatial context in the low-level features for final prediction. Our experiments show that the high-order statistics significantly boost the performance on confusing objects. Our method achieves competitive performance without bells and whistles on three benchmarks, i.e, Cityscapes, ADE20K and Pascal-Context with the mIoU of 81.6%, 45.3% and 52.9%.
20.Multi-Task Learning from Videos via Efficient Inter-Frame Attention ⬇️
Prior work in multi-task learning has mainly focused on predictions on a single image. In this work, we present a new approach for multi-task learning from videos. Our approach contains a novel inter-frame attention module which allows learning of task-specific attention across frames. We embed the attention module in a "slow-fast" architecture, where the slower network runs on sparsely sampled keyframes and the lightweight shallow network runs on non-key frames at a high frame rate. We further propose an effective adversarial learning strategy to encourage the slow and fast network to learn similar features. The proposed architecture ensures low-latency multi-task learning while maintaining high quality prediction. Experiments show competitive accuracy compared to state-of-the-art on two multi-task learning benchmarks while reducing the number of floating point operations (FLOPs) by 70%. Meanwhile, our attention based feature propagation outperforms other feature propagation methods in accuracy by up to 90% reduction of FLOPs.
21.Constraining Temporal Relationship for Action Localization ⬇️
Recently, temporal action localization (TAL), i.e., finding specific action segments in untrimmed videos, has attracted increasing attentions of the computer vision community. State-of-the-art solutions for TAL involves predicting three values at each time point, corresponding to the probabilities that the action starts, continues and ends, and post-processing these curves for the final localization. This paper delves deep into this mechanism, and argues that existing approaches mostly ignored the potential relationship of these curves, and results in low quality of action proposals. To alleviate this problem, we add extra constraints to these curves, e.g., the probability of ''action continues'' should be relatively high between probability peaks of ''action starts'' and ''action ends'', so that the entire framework is aware of these latent constraints during an end-to-end optimization process. Experiments are performed on two popular TAL datasets, THUMOS14 and ActivityNet1.3. Our approach clearly outperforms the baseline both quantitatively (in terms of the AR@AN and mAP) and qualitatively (the curves in the testing stage become much smoother). In particular, when we build our constraints beyond TSA-Net and PGCN, we achieve the state-of-the-art performance especially at strict high IoU settings. The code will be available.
22.Restricted Structural Random Matrix for Compressive Sensing ⬇️
Compressive sensing (CS) is well-known for its unique functionalities of sensing, compressing, and security (i.e. CS measurements are equally important). However, there is a tradeoff. Improving sensing and compressing efficiency with prior signal information tends to favor particular measurements, thus decrease the security. This work aimed to improve the sensing and compressing efficiency without compromise the security with a novel sampling matrix, named Restricted Structural Random Matrix (RSRM). RSRM unified the advantages of frame-based and block-based sensing together with the global smoothness prior (i.e. low-resolution signals are highly correlated). RSRM acquired compressive measurements with random projection (equally important) of multiple randomly sub-sampled signals, which was restricted to be the low-resolution signals (equal in energy), thereby, its observations are equally important. RSRM was proven to satisfies the Restricted Isometry Property and shows comparable reconstruction performance with recent state-of-the-art compressive sensing and deep learning-based methods.
23.3D Gated Recurrent Fusion for Semantic Scene Completion ⬇️
This paper tackles the problem of data fusion in the semantic scene completion (SSC) task, which can simultaneously deal with semantic labeling and scene completion. RGB images contain texture details of the object(s) which are vital for semantic scene understanding. Meanwhile, depth images capture geometric clues of high relevance for shape completion. Using both RGB and depth images can further boost the accuracy of SSC over employing one modality in isolation. We propose a 3D gated recurrent fusion network (GRFNet), which learns to adaptively select and fuse the relevant information from depth and RGB by making use of the gate and memory modules. Based on the single-stage fusion, we further propose a multi-stage fusion strategy, which could model the correlations among different stages within the network. Extensive experiments on two benchmark datasets demonstrate the superior performance and the effectiveness of the proposed GRFNet for data fusion in SSC. Code will be made available.
24.Dual-Attention GAN for Large-Pose Face Frontalization ⬇️
Face frontalization provides an effective and efficient way for face data augmentation and further improves the face recognition performance in extreme pose scenario. Despite recent advances in deep learning-based face synthesis approaches, this problem is still challenging due to significant pose and illumination discrepancy. In this paper, we present a novel Dual-Attention Generative Adversarial Network (DA-GAN) for photo-realistic face frontalization by capturing both contextual dependencies and local consistency during GAN training. Specifically, a self-attention-based generator is introduced to integrate local features with their long-range dependencies yielding better feature representations, and hence generate faces that preserve identities better, especially for larger pose angles. Moreover, a novel face-attention-based discriminator is applied to emphasize local features of face regions, and hence reinforce the realism of synthetic frontal faces. Guided by semantic segmentation, four independent discriminators are used to distinguish between different aspects of a face (\ie skin, keypoints, hairline, and frontalized face). By introducing these two complementary attention mechanisms in generator and discriminator separately, we can learn a richer feature representation and generate identity preserving inference of frontal views with much finer details (i.e., more accurate facial appearance and textures) comparing to the state-of-the-art. Quantitative and qualitative experimental results demonstrate the effectiveness and efficiency of our DA-GAN approach.
25.Multilinear Compressive Learning with Prior Knowledge ⬇️
The recently proposed Multilinear Compressive Learning (MCL) framework combines Multilinear Compressive Sensing and Machine Learning into an end-to-end system that takes into account the multidimensional structure of the signals when designing the sensing and feature synthesis components. The key idea behind MCL is the assumption of the existence of a tensor subspace which can capture the essential features from the signal for the downstream learning task. Thus, the ability to find such a discriminative tensor subspace and optimize the system to project the signals onto that data manifold plays an important role in Multilinear Compressive Learning. In this paper, we propose a novel solution to address both of the aforementioned requirements, i.e., How to find those tensor subspaces in which the signals of interest are highly separable? and How to optimize the sensing and feature synthesis components to transform the original signals to the data manifold found in the first question? In our proposal, the discovery of a high-quality data manifold is conducted by training a nonlinear compressive learning system on the inference task. Its knowledge of the data manifold of interest is then progressively transferred to the MCL components via multi-stage supervised training with the supervisory information encoding how the compressed measurements, the synthesized features, and the predictions should be like. The proposed knowledge transfer algorithm also comes with a semi-supervised adaption that enables compressive learning models to utilize unlabeled data effectively. Extensive experiments demonstrate that the proposed knowledge transfer method can effectively train MCL models to compressively sense and synthesize better features for the learning tasks with improved performances, especially when the complexity of the learning task increases.
26.The Tree Ensemble Layer: Differentiability meets Conditional Computation ⬇️
Neural networks and tree ensembles are state-of-the-art learners, each with its unique statistical and computational advantages. We aim to combine these advantages by introducing a new layer for neural networks, composed of an ensemble of differentiable decision trees (a.k.a. soft trees). While differentiable trees demonstrate promising results in the literature, in practice they are typically slow in training and inference as they do not support conditional computation. We mitigate this issue by introducing a new sparse activation function for sample routing, and implement true conditional computation by developing specialized forward and backward propagation algorithms that exploit sparsity. Our efficient algorithms pave the way for jointly training over deep and wide tree ensembles using first-order methods (e.g., SGD). Experiments on 23 classification datasets indicate over 10x speed-ups compared to the differentiable trees used in the literature and over 20x reduction in the number of parameters compared to gradient boosted trees, while maintaining competitive performance. Moreover, experiments on CIFAR, MNIST, and Fashion MNIST indicate that replacing dense layers in CNNs with our tree layer reduces the test loss by 7-53% and the number of parameters by 8x. We provide an open-source TensorFlow implementation with a Keras API.
27.Learning Bijective Feature Maps for Linear ICA ⬇️
Separating high-dimensional data like images into independent latent factors remains an open research problem. Here we develop a method that jointly learns a linear independent component analysis (ICA) model with non-linear bijective feature maps. By combining these two methods, ICA can learn interpretable latent structure for images. For non-square ICA, where we assume the number of sources is less than the dimensionality of data, we achieve better unsupervised latent factor discovery than flow-based models and linear ICA. This performance scales to large image datasets such as CelebA.
28.Deep Learning in Medical Ultrasound Image Segmentation: a Review ⬇️
Applying machine learning technologies, especially deep learning, into medical image segmentation is being widely studied because of its state-of-the-art performance and results. It can be a key step to provide a reliable basis for clinical diagnosis, such as 3D reconstruction of human tissues, image-guided interventions, image analyzing and visualization. In this review article, deep-learning-based methods for ultrasound image segmentation are categorized into six main groups according to their architectures and training at first. Secondly, for each group, several current representative algorithms are selected, introduced, analyzed and summarized in detail. In addition, common evaluation methods for image segmentation and ultrasound image segmentation datasets are summarized. Further, the performance of the current methods and their evaluations are reviewed. In the end, the challenges and potential research directions for medical ultrasound image segmentation are discussed.
29.Robust Quantization: One Model to Rule Them All ⬇️
Neural network quantization methods often involve simulating the quantization process during training. This makes the trained model highly dependent on the precise way quantization is performed. Since low-precision accelerators differ in their quantization policies and their supported mix of data-types, a model trained for one accelerator may not be suitable for another. To address this issue, we propose KURE, a method that provides intrinsic robustness to the model against a broad range of quantization implementations. We show that KURE yields a generic model that may be deployed on numerous inference accelerators without a significant loss in accuracy.
30.Image Entropy for Classification and Analysis of Pathology Slides ⬇️
Pathology slides of lung malignancies are classified using the "Salient Slices" technique described in Frank et al., 2020. A four-fold cross-validation study using a small image set (42 adenocarcinoma slides and 42 squamous cell carcinoma slides) produced fully correct classifications in each fold. Probability maps enable visualization of the underlying basis for a classification.
31.Deflecting Adversarial Attacks ⬇️
There has been an ongoing cycle where stronger defenses against adversarial attacks are subsequently broken by a more advanced defense-aware attack. We present a new approach towards ending this cycle where we "deflect'' adversarial attacks by causing the attacker to produce an input that semantically resembles the attack's target class. To this end, we first propose a stronger defense based on Capsule Networks that combines three detection mechanisms to achieve state-of-the-art detection performance on both standard and defense-aware attacks. We then show that undetected attacks against our defense often perceptually resemble the adversarial target class by performing a human study where participants are asked to label images produced by the attack. These attack images can no longer be called "adversarial'' because our network classifies them the same way as humans do.
32.Picking Winning Tickets Before Training by Preserving Gradient Flow ⬇️
Overparameterization has been shown to benefit both the optimization and generalization of neural networks, but large networks are resource hungry at both training and test time. Network pruning can reduce test-time resource requirements, but is typically applied to trained networks and therefore cannot avoid the expensive training process. We aim to prune networks at initialization, thereby saving resources at training time as well. Specifically, we argue that efficient training requires preserving the gradient flow through the network. This leads to a simple but effective pruning criterion we term Gradient Signal Preservation (GraSP). We empirically investigate the effectiveness of the proposed method with extensive experiments on CIFAR-10, CIFAR-100, Tiny-ImageNet and ImageNet, using VGGNet and ResNet architectures. Our method can prune 80% of the weights of a VGG-16 network on ImageNet at initialization, with only a 1.6% drop in top-1 accuracy. Moreover, our method achieves significantly better performance than the baseline at extreme sparsity levels.
33.Evolutionary Optimization of Deep Learning Activation Functions ⬇️
The choice of activation function can have a large effect on the performance of a neural network. While there have been some attempts to hand-engineer novel activation functions, the Rectified Linear Unit (ReLU) remains the most commonly-used in practice. This paper shows that evolutionary algorithms can discover novel activation functions that outperform ReLU. A tree-based search space of candidate activation functions is defined and explored with mutation, crossover, and exhaustive search. Experiments on training wide residual networks on the CIFAR-10 and CIFAR-100 image datasets show that this approach is effective. Replacing ReLU with evolved activation functions results in statistically significant increases in network accuracy. Optimal performance is achieved when evolution is allowed to customize activation functions to a particular task; however, these novel activation functions are shown to generalize, achieving high performance across tasks. Evolutionary optimization of activation functions is therefore a promising new dimension of metalearning in neural networks.
34.AIBench: An Agile Domain-specific Benchmarking Methodology and an AI Benchmark Suite ⬇️
Domain-specific software and hardware co-design is encouraging as it is much easier to achieve efficiency for fewer tasks. Agile domain-specific benchmarking speeds up the process as it provides not only relevant design inputs but also relevant metrics, and tools. Unfortunately, modern workloads like Big data, AI, and Internet services dwarf the traditional one in terms of code size, deployment scale, and execution path, and hence raise serious benchmarking challenges.
This paper proposes an agile domain-specific benchmarking methodology. Together with seventeen industry partners, we identify ten important end-to-end application scenarios, among which sixteen representative AI tasks are distilled as the AI component benchmarks. We propose the permutations of essential AI and non-AI component benchmarks as end-to-end benchmarks. An end-to-end benchmark is a distillation of the essential attributes of an industry-scale application. We design and implement a highly extensible, configurable, and flexible benchmark framework, on the basis of which, we propose the guideline for building end-to-end benchmarks, and present the first end-to-end Internet service AI benchmark.
The preliminary evaluation shows the value of our benchmark suite---AIBench against MLPerf and TailBench for hardware and software designers, micro-architectural researchers, and code developers. The specifications, source code, testbed, and results are publicly available from the web site \url{this http URL}.