For any recognition system, the ability to identify novel class samples during inference is an important aspect of the system’s robustness. This problem of detecting novel class samples during inference is commonly referred to as Multiple Class Novelty Detection. In this paper, we propose a novel method that makes deep convolutional neural networks robust to novel classes. Specifically, during training one branch performs traditional classification (referred to as global inference), and the other branch provides patch-level information to keep track of the class-specific activation patterns (referred to as local inference). Both global and local branch information are combined to train a novelty detection network, which is used during inference to identify novel classes. We evaluate the proposed method on four datasets (Caltech256, CUB-200, Stanford Dogs and FounderType-200) and show that the proposed method is able to identify novel class samples better compared to the other deep convolutional neural network based methods.
论文检索
输入标题、作者或关键词，从 1,358 篇学术成果中精准定位
Real-time 3D object detection is crucial for autonomous cars. Achieving promising performance with high efficiency, voxel-based approaches have received considerable attention. However, previous methods model the input space with features extracted from equally divided sub-regions without considering that point cloud is generally non-uniformly distributed over the space. To address this issue, we propose a novel 3D object detection framework with dynamic information modeling. The proposed framework is designed in a coarse-to-fine manner. Coarse predictions are generated in the first stage via a voxel-based region proposal network. We introduce InfoFocus, which improves the coarse detections by adaptively refining features guided by the information of point cloud density. Experiments are conducted on the large-scale nuScenes 3D detection benchmark. Results show that our framework achieves the state-of-the-art performance with 31 FPS and improves our baseline significantly by 9.0% mAP on the nuScenes test set.
Spiking Neural Networks (SNNs) are promising for enabling low-power event-driven data analytics. The best performing SNNs for image recognition tasks are obtained by converting a trained deep learning Analog Neural Network (ANN) composed of Rectified Linear Unit (ReLU) activation to SNN consisting of Integrate-and-Fire (IF) neurons with ""proper"" firing thresholds. However, this has come at the cost of accuracy loss and higher inference latency due to lack of a notion of time. In this work, we propose an ANN to SNN conversion methodology that uses a time-based coding scheme, named Temporal-Switch-Coding (TSC), and a corresponding TSC spiking neuron model. Each input image pixel is presented using two spikes and the timing between the two spiking instants is proportional to the pixel intensity. The real-valued ReLU activations in ANN are encoded using the spike-times of the TSC neurons in the converted TSC-SNN. At most two memory accesses and two addition operations are performed for each synapse during the whole inference, which significantly improves the SNN energy efficiency. We demonstrate the proposed TSC-SNN for VGG-16, ResNet-20 and ResNet-34 SNNs on datasets including CIFAR-10 (93.63% top-1), CIFAR-100 (70.97% top-1) and ImageNet (73.46% top-1 accuracy). It surpasses the best inference accuracy of the converted rate-encoded SNN with 7-14.5 times lesser inference latency, and 30-60 times fewer addition operations and memory accesses per inference across datasets.
For safety-critical applications such as autonomous driving, CNNs have to be robust with respect to unavoidable image corruptions, such as image noise. While previous works addressed the task of robust prediction in the context of full-image classification, we consider it for dense semantic segmentation. We build upon an insight from image classification that output robustness can be improved by increasing the network-bias towards object shapes. We present a new training schema that increases this shape bias. Our basic idea is to alpha-blend a portion of the RGB training images with faked images, where each class-label is given a fixed, randomly chosen color that is not likely to appear in real imagery. This forces the network to rely more strongly on shape cues. We call this data augmentation technique ""Painting-by-Numbers''. We demonstrate the effectiveness of our training schema for DeepLabv3$+$ with various network backbones, MobileNet-V2, ResNets, and Xception, and evaluate it on the Cityscapes dataset. With respect to our 16 different types of image corruptions and 5 different network backbones, we are in 74% better than training with clean data. For cases where we are worse than a model trained without our training schema, it is mostly only marginally worse. However, for some image corruptions such as images with noise, we see a considerable performance gain of up to 25%.
For semantic segmentation, most existing real-time deep mod-els trained with each frame independently may produce inconsistent results when tested on a video sequence. A few methods take the correlations in the video sequence into account, e.g., by propagating the results to the neighboring frames using optical flow or extracting frame representations using multi-frame information, which may lead to inaccurate results or unbalanced latency. In contrast, here we explicitly consider the temporal consistency among frames as extra constraints during training and process each frame independently in the inference phase. Thus no computation overhead is introduced for inference. Compact models are employed for real-time execution. To narrow the performance gap between compact models and large models, new temporal knowledge distillation methods are designed. Weighing among accuracy, temporal smoothness, and efficiency, our proposed method outperforms previous keyframe based methods and corresponding baselines which are trained with each frame independently on benchmark datasets including cityscapes and Camvid.
Video super-resolution (VSR) aims to utilize multiple low-resolution frames to generate a high-resolution prediction for each frame. In this process, inter- and intra-frames are the key sources for exploiting temporal and spatial information. However, there are a couple of limitations for existing VSR methods. First, optical flow is often used to establish one-on-one temporal correspondences. But flow estimation itself is error-prone and hence largely affects the ultimate recovery result. Second, similar patterns existing in natural images are rarely exploited for the VSR task. Motivated by these findings, we propose a temporal multi-correspondence aggregation strategy to leverage most similar patches across frames, and also a cross-scale nonlocal-correspondence aggregation scheme to explore self-similarity of images across scales. Based on these two novel modules, we build an effective multi-correspondence aggregation network (MuCAN) for VSR. Our method achieves state-of-the-art results on multiple benchmark datasets. Extensive experiments justify the effectiveness of our method.
Graph convolutional networks (GCNs) have been applied to 3D human pose estimation (HPE) from 2D body joint detections and have shown encouraging performance. One limitation of the vanilla graph convolution is that it models the relationships between neighboring nodes via a shared weight matrix. This is suboptimal for articulated body modeling as the relations between different body joints are different. The objective of this paper is to have a comprehensive and systematic study of weight sharing in GCNs for 3D HPE. We first show there are two different ways to interpret a GCN depending on whether feature transformation occurs before or after feature aggregation. These two interpretations lead to five different weight sharing methods, and three more variants can be derived by decoupling the self-connections with other edges. We conduct extensive ablation study on these weight sharing methods under controlled settings and obtain new conclusions that will benefit the community.
Vision-and-Language Navigation (VLN) is unique in that it requires turning relatively general natural-language instructions into robot agent actions, on the basis of visible environments. This requires to extract value from two very different types of natural-language information. The first is object description (e.g., `table', `door'), each presenting as a tip for the agent to determine the next action by finding the item visible in the environment, and the second is action specification (e.g., `go straight', `turn left') which allows the robot to directly predict the next movements without relying on visual perceptions. However, most existing methods pay few attention to distinguish these information from each other during instruction encoding and mix together the matching between textual object/action encoding and visual perception/orientation features of candidate viewpoints. In this paper, we propose an Object-and-Action Aware Model (OAAM) that processes these two different forms of natural language based instruction separately. This enables each process to match object-centered/action-centered instruction to their own counterpart visual perception/action orientation flexibly. However, one side-issue caused by above solution is that an object mentioned in instructions may be observed in the direction of two or more candidate viewpoints, thus the OAAM may not predict the viewpoint on the shortest path as the next action. To handle this problem, we design a simple but effective path loss to penalize trajectories deviating from the ground truth path. Experimental results demonstrate the effectiveness of the proposed model and path loss, and the superiority of their combination with a 50% SPL score on the R2R dataset and a 40% CLS score on the R4R dataset in unseen environments, outperforming the previous state-of-the-art.
In this work we consider the problem of learning a classifier from noisy labels when a few clean labeled examples are given. The structure of clean and noisy data is modeled by a graph per class and Graph Convolutional Networks (GCN) are used to predict class relevance of noisy examples. For each class, the GCN is treated as a binary classifier, which learns to discriminate clean from noisy examples using a weighted binary cross-entropy loss function. The GCN-inferred ""clean"" probability is then exploited as a relevance measure. Each noisy example is weighted by its relevance when learning a classifier for the end task. We evaluate our method on an extended version of a few-shot learning problem, where the few clean examples of novel classes are supplemented with additional noisy data. Experimental results show that our GCN-based cleaning process significantly improves the classification accuracy over not cleaning the noisy data, as well as standard few-shot classification where only few clean examples are used.
Can we automatically group images into semantically meaningful clusters when ground-truth annotations are absent? The task of unsupervised image classification remains an important, and open challenge in computer vision. Several recent approaches have tried to tackle this problem in an end-to-end fashion. In this paper, we deviate from recent works, and advocate a two-step approach where feature learning and clustering are decoupled. First, a self-supervised task from representation learning is employed to obtain semantically meaningful features. Second, we use the obtained features as a prior in a learnable clustering approach. In doing so, we remove the ability for cluster learning to depend on low-level features, which is present in current end-to-end learning approaches. Experimental evaluation shows that we outperform state-of-the-art methods by large margins, in particular +26.6% on CIFAR10, +25.0% on CIFAR100-20 and +21.3% on STL10 in terms of classification accuracy. Furthermore, our method is the first to perform well on a large-scale dataset for image classification. In particular, we obtain promising results on ImageNet, and outperform several semi-supervised learning methods in the low-data regime without the use of any ground-truth annotations.
Graph matching has been a fundamental problem in computer vision and pattern recognition, for its practical flexibility as well as NP hardness challenge. Though the matching between two graphs and among multiple graphs have been intensively studied in literature, the online setting for incremental matching of a stream of graphs has been rarely considered. In this paper, we treat the graphs as graphs on a super-graph, and propose a novel breadth first search based method for expanding the neighborhood on the super-graph for a new coming graph, such that the matching with the new graph can be efficiently performed within the constructed neighborhood. Then depth first search is performed to update the overall pairwise matchings. Moreover, we show our approach can also be readily used in the batch mode setting, by adaptively determining the order of coming graph batch for matching, still under the neighborhood expansion based incremental matching framework. Extensive experiments on both online and offline matching of graph collections show our approach’s state-of-the-art accuracy and efficiency. The source code will be made public available.
Thanks for Nothing: Predicting Zero-Valued Activations with Lightweight Convolutional Neural Networks
PDF ↗Convolutional neural networks (CNNs) introduce state-of-the-art results for various tasks with the price of high computational demands. Inspired by the observation that spatial correlation exists in CNN output feature maps (ofms), we propose a method to dynamically predict whether ofm activations are zero-valued or not according to their neighboring activation values, thereby avoiding zero-valued activations and reducing the number of convolution operations. We implement the zero activation predictor (ZAP) with a lightweight CNN, which imposes negligible overheads and is easy to deploy on existing models. ZAPs are trained by mimicking hidden layer ouputs; thereby, enabling a parallel and label-free training. Furthermore, without retraining, each ZAP can be tuned to a different operating point trading accuracy for MAC reduction.
In this paper we propose a rotation-invariant deep network for point clouds analysis. Point-based deep networks are commonly designed to recognize roughly aligned 3D shapes based on point coordinates, but suffer from performance drops with shape rotations. Some geometric features, e.g., distances and angles of points as inputs of network, are rotation-invariant but lose positional information of points. In this work, we propose a novel deep network for point clouds by incorporating positional information of points as inputs while yielding rotation-invariance. The network is hierarchical and relies on two modules: a positional feature embedding block and a relational feature embedding block. Both modules and the whole network are proven to be rotation-invariant when processing point clouds as input. Experiments show state-of-the-art classification and segmentation performances on benchmark datasets, and ablation studies demonstrate effectiveness of the network design.
Domain generalization models learn to generalize to previously unseen domains, but suffer from prediction uncertainty and domain shift. In this paper, we address both problems. We introduce a probabilistic meta-learning model for domain generalization, in which classifier parameters shared across domains are modeled as distributions. This enables better handling of prediction uncertainty on unseen domains. To deal with domain shift, we learn domain-invariant representations by the proposed principle of meta variational information bottleneck, we call MetaVIB. MetaVIB is derived from novel variational bounds of mutual information, by leveraging the meta-learning setting of domain generalization. Through episodic training, MetaVIB learns to gradually narrow domain gaps to establish domain-invariant representations, while simultaneously maximizing prediction accuracy. We conduct experiments on three benchmarks for cross-domain visual recognition. Comprehensive ablation studies validate the benefits of MetaVIB for domain generalization. The comparison results demonstrate our method outperforms previous approaches consistently.
Recent studies have shown that DNNs can be compromised by backdoor attacks crafted at training time. A backdoor attack installs a backdoor into the victim model by injecting a backdoor pattern into a small set of training examples. The victim model behaves normally on clean test data, yet consistently predicts a specific (likely incorrect) target class whenever the backdoor pattern is present in an input example. While existing backdoor attacks are effective, they are not stealthy. The modifications made on training data or labels are often suspicious and can be easily detected by simple data filtering or human inspection. In this paper, we present a new type of backdoor attack inspired by an important natural phenomenon: reflection. Using mathematical modeling of physical reflection models, we propose reflection backdoor (Refool) to plant reflections as backdoor into a victim model. We demonstrate on 3 computer vision tasks and 5 datasets that, Refool can attack state-of-the-art DNNs with high success rate, and is more resistant to state-of-the-art backdoor defenses than existing attacks.
We formulate counting as a sequential decision problem and present a novel crowd counting model solvable by deep reinforcement learning. In contrast to existing counting models that directly output count values, we divide one-step estimation into a sequence of much easier and more tractable sub-decision problems. Such sequential decision nature corresponds exactly to a physical process in reality|scale weighing. Inspired by scale weighing, we propose a novel ‘counting scale’ termed LibraNet where the count value is analogized by weight. By virtually placing a crowd image on one side of a scale, LibraNet (agent) sequentially learns to place appropriate weights on the other side to match the crowd count. At each step, LibraNet chooses one weight (action) from the weight box (the pre-defined action pool) according to the current crowd image features and weights placed on the scale pan (state). LibraNet is required to learn to balance the scale according to the feedback of the needle (Q values). We show that LibraNet exactly implements scale weighing by visualizing the decision process how LibraNet chooses actions. Extensive experiments demonstrate the effectiveness of our design choices and report state-of-the-art results on a few crowd counting benchmarks, including ShanghaiTech, UCF_CC_50 and UCF-QNRF. We also demonstrate good cross-dataset generalization of LibraNet. Code and models are made available at https://git.io/libranet
Take an Emotion Walk: Perceiving Emotions from Gaits Using Hierarchical Attention Pooling and Affective Mapping
PDF ↗We present an autoencoder-based semi-supervised approach to classify perceived human emotions from walking styles obtained from videos or motion-captured data and represented as sequences of 3D poses. Given the motion on each joint in the pose at each time step extracted from 3D pose sequences, we hierarchically pool these joint motions in a bottom-up manner in the encoder, following the kinematic chains in the human body. We also constrain the latent embeddings of the encoder to contain the space of psychologically-motivated affective features underlying the gaits. We train the decoder to reconstruct the motions per joint per time step in a top-down manner from the latent embeddings. For the annotated data, we also train a classifier to map the latent embeddings to emotion labels. Our semi-supervised approach achieves a mean average precision of 0.84 on the Emotion-Gait benchmark dataset, which contains both labeled and unlabeled gaits collected from multiple sources. We outperform current state-of-art algorithms for both emotion recognition and action recognition from 3D gaits by 7%-23% on the absolute. More importantly, we improve the average precision by 10%-50% on the absolute on classes that each makes up less than 25% of the labeled part of the Emotion-Gait benchmark dataset.
Though many successful approaches have been proposed for recognizing events in short and homogeneous videos, doing so with long and complex videos remains a challenge. One particular reason is that events in long and complex videos can consist of multiple heterogeneous sub-activities (in terms of rhythms, activity variants, composition order, etc.) within quite a long period. This fact brings about two main difficulties: excessive/varying length and complex video dynamic/rhythm. To address this, we propose Rhythmic RNN (RhyRNN) which is capable of handling long video sequences (up to 3,000 frames) as well as capturing rhythms at different scales. We also propose two novel modules: diversity-driven pooling (DivPool) and bilinear reweighting (BR), which consistently and hierarchically abstract higher-level information. We study the behavior of RhyRNN and empirically show that our method works well even when mph{only event-level labels are available} in the training stage (compared to algorithms requiring sub-activity labels for recognition), and thus is more practical when the sub-activity labels are missing or difficult to obtain. Extensive experiments on several public datasets demonstrate that, even mph{without fine-tuning the feature backbones}, our method can achieve promising performance for long and complex videos that contain multiple sub-activities.
Defocus blur arises in images that are captured with a shallow depth of field due to the use of a wide aperture. Correcting defocus blur is challenging because the blur is spatially varying and difficult to estimate. We propose an effective defocus deblurring method that exploits data available on dual-pixel (DP) sensors found on most modern cameras. DP sensors are used to assist a camera's auto-focus by capturing two sub-aperture views of the scene in a single image shot. The two sub-aperture images are used to calculate the appropriate lens position to focus on a particular scene region and are discarded afterwards. We introduce a deep neural network (DNN) architecture that uses these discarded sub-aperture images to reduce defocus blur. A key contribution of our effort is a carefully captured dataset of 500 scenes (2000 images) where each scene has: (i) an image with defocus blur captured at a large aperture; (ii) the two associated DP sub-aperture views; and (iii) the corresponding all-in-focus image captured with a small aperture. Our proposed DNN produces results that are significantly better than conventional single image methods in terms of both quantitative and perceptual metrics -- all from data that is already available on the camera but ignored.
To achieve robustness in Re-Identification, standard methods leverage tracking information in a Video-To-Video fashion. However, these solutions face a large drop in performance for single image queries (e.g., Image-To-Video setting). Recent works address this severe degradation by transferring temporal information from a Video-based network to an Image-based one. In this work, we devise a training strategy that allows the transfer of a superior knowledge, arising from a set of views depicting the target object. Our proposal - Views Knowledge Distillation (VKD) - pins this visual variety as a supervision signal within a teacher-student framework, where the teacher educates a student who observes fewer views. As a result, the student outperforms not only its teacher but also the current state-of-the-art in Image-To-Video by a wide margin (6.3% mAP on MARS, 8.6% on Duke-Video-ReId and 5% on VeRi-776). A thorough analysis - on Person, Vehicle and Animal Re-ID - investigates the properties of VKD from a qualitatively and quantitatively perspective. Code is available at https://github.com/aimagelab/VKD.