People learn throughout life. However, incrementally updating conventional neural networks leads to catastrophic forgetting. A common remedy is replay, which is inspired by how the brain consolidates memory. Replay involves fine-tuning a network on a mixture of new and old instances. While there is neuroscientific evidence that the brain replays compressed memories, existing methods for convolutional networks replay raw images. Here, we propose REMIND, a brain-inspired approach that enables efficient replay with compressed representations. REMIND is trained in an online manner, meaning it learns one example at a time, which is closer to how humans learn. Under the same constraints, REMIND outperforms other methods for incremental class learning on the ImageNet ILSVRC-2012 dataset. We probe REMIND's robustness to data ordering schemes known to induce catastrophic forgetting. We demonstrate REMIND's generality by pioneering online learning for Visual Question Answering (VQA).
论文检索
输入标题、作者或关键词，从 8,216 篇学术成果中精准定位
Photorealistic style transfer is the task of transferring the artistic style of an image onto a content target, producing a result that is plausibly taken with a camera. Recent approaches, based on deep neural networks, produce impressive results but are either too slow to run at practical resolutions, or still contain objectionable artifacts. We propose a new end-to-end model for photorealistic style transfer that is both fast and inherently generates photorealistic results. The core of our approach is a feed-forward neural network that learns local edge-aware affine transforms that automatically obey the photorealism constraint. When trained on a diverse set of images and a variety of styles, our model can robustly apply style transfer to an arbitrary pair of input images. Compared to the state of the art, our method produces visually superior results and is three orders of magnitude faster, enabling real-time performance at 4K on a mobile phone. We validate our method with ablation and user studies.
Long-tail recognition tackles the natural non-uniformly distributed data in real-world scenarios. While modern classifiers perform well on populated classes, its performance degrades significantly on tail classes. Humans, however, are less affected by this since, when confronted with uncertain examples, they simply opt to provide coarser predictions. Motivated by this, a deep realistic taxonomic classifier (Deep-RTC) is proposed as a new solution to the long-tail problem, combining realism with hierarchical predictions. The model has the option to reject classifying samples at different levels of the taxonomy, once it cannot guarantee the desired performance. Deep-RTC is implemented with a stochastic tree sampling during training to simulate all possible classification conditions at finer or coarser levels and a rejection mechanism at inference time. Experiments on the long-tailed version of four datasets, CIFAR100, AWA2, Imagenet, and iNaturalist, demonstrate that the proposed approach preserves more information on all classes with different popularity levels. Deep-RTC also outperforms the state-of-the-art methods in longtailed recognition, hierarchical classification, and learning with rejection literature using the proposed correctly predicted bits (CPB) metric.
We propose a method to train a model so it can learn new classification tasks while improving with each task solved. This amounts to combining meta-learning with incremental learning. Different tasks can have disjoint classes, so one cannot directly align different classifiers as done in model distillation. On the other hand, simply aligning features shared by all classes does not allow the base model sufficient flexibility to evolve to solve new tasks. We therefore indirectly align features relative to a minimal set of anchor classes. Such indirect discriminant alignment (IDA) adapts a new model to old classes without the need to re-process old data, while leaving maximum flexibility for the model to adapt to new tasks. This process enables incrementally improving the model by processing multiple learning episodes, each representing a different learning task, even with few training examples. Experiments on few-shot learning benchmarks show that this incremental approach performs favorably compared to training the model with the entire dataset at once.
We introduce a new problem of mph{retrieving} 3D models that are mph{deformable} to a given query shape and present a novel deep mph{deformation-aware} embedding to solve this retrieval task. 3D model retrieval is a fundamental operation for recovering a clean and complete 3D model from a noisy and partial 3D scan. However, given a finite collection of 3D shapes, even the closest model to a query may not be satisfactory. This motivates us to apply 3D model deformation techniques to adapt the retrieved model so as to better fit the query. Yet, certain restrictions are enforced in most 3D deformation techniques to preserve important features of the original model that prevent a perfect fitting of the deformed model to the query. This gap between the deformed model and the query induces mph{asymmetric} relationships among the models, which cannot be handled by typical metric learning techniques. Thus, to retrieve the best models for fitting, we propose a novel deep embedding approach that learns the asymmetric relationships by leveraging location-dependent egocentric distance fields. We also propose two strategies for training the embedding network. We demonstrate that both of these approaches outperform other baselines in our experiments with both synthetic and real data. Our project page can be found at \href{https://deformscan2cad.github.io/}{deformscan2cad.github.io}.
In this work, we propose a new unsupervised image segmentation approach based on mutual information maximization between different constructed views of the inputs. Taking inspiration from autoregressive generative models, that predict the current pixel from past pixels in a raster-scan ordering created with masked convolutions, we propose different autoregressive orderings over the inputs using various forms of masked convolutions to construct different views of the data. For a given input, the model produces a pair of predictions with two valid orderings, and is then trained to maximize the mutual information between the two outputs. These outputs can either be low-dimensional features for representation learning or output clusters corresponding to semantic labels for clustering. While masked convolutions are used during training, in inference, no masking is applied and we fall back to the standard convolution where the model has access to the full input. The proposed method outperforms the current state-of-the-art on unsupervised image segmentation. It is simple and easy to implement, and can be extended to other visual tasks and integrated seamlessly into existing unsupervised learning methods requiring different views of the data.
When a very fast dynamic event is recorded with a low-framerate camera, the resulting video suffers from severe motion blur (due to exposure time) and motion aliasing (due to low sampling rate in time). True Temporal Super-Resolution (TSR) is more than just Temporal-Interpolation (increasing framerate). It can also recover new high temporal frequencies beyond the temporal Nyquist limit of the input video, thus resolving both motion-blur and motion aliasing – effects that temporal frame interpolation (as sophisticated as it may be) cannot undo. In this paper we propose a “Deep Internal Learning” approach for true TSR. We train a video-specific FCN on examples extracted directly from the low-framerate input video. Our method exploits the strong recurrence of small space-time patches inside a single video sequence, both within and across different spatio-temporal scales of the video.We further observe (for the first time) that small space-time patches recur also across-dimensions of the video sequence – i.e., by swapping the spatial and temporal dimensions. In particular, the higher spatial resolution of video frames provides strong examples as to how to increase the temporal resolution of that video. Such internal video-specific examples give rise to strong self-supervision, requiring no data but the input video itself. This results in Zero-Shot Temporal-SR of complex videos, which removes both motion blur and motion aliasing, outperforming previous supervised methods trained on external video datasets.
State-of-the-art navigation methods leverage a spatial memory to generalize to new environments, but their occupancy maps are limited to capturing the geometric structures directly observed by the agent. We propose occupancy anticipation, where the agent uses its egocentric RGB-D observations to infer the occupancy state beyond the visible regions. In doing so, the agent builds its spatial awareness more rapidly, which facilitates efficient exploration and navigation in 3D environments. By exploiting context in both the egocentric views and top-down maps our model successfully anticipates a broader map of the environment, with performance significantly better than strong baselines. Furthermore, when deploying our model for the sequential decision-making tasks of exploration and navigation, we outperform state-of-the-art methods on the Gibson and Matterport3D datasets. Our approach is the winning entry in the 2020 Habitat PointNav Challenge. Project page: http://vision.cs.utexas.edu/projects/occupancy_anticipation
Data from new categories are continuously being discovered, which has sparked significant amount of research in developing approaches which generalizes to previously unseen categories, i.e. zero-shot setting. Zero-shot sketch-based image retrieval~(ZS-SBIR) is one such problem in the context of cross-domain retrieval, which has received lot of attention due to its various real-life applications. Since most real-world training data have a fair amount of imbalance; in this work, for the first time in literature, we extensively study the effect of training data imbalance on the generalization to unseen categories, with ZS-SBIR as the application area. We evaluate several state-of-the-art data imbalance mitigating techniques and analyze their results. Furthermore, we propose a novel framework AMDReg (Adaptive Margin Diversity Regularizer), which ensures that the embeddings of the sketch and images in the latent space are not only semantically meaningful, but they also are separated according to their class-representations in the training set. The proposed approach is model-independent, and it can be incorporated seamlessly with several state-of-the-art ZS-SBIR methods to improve their performance under imbalanced condition. Extensive experiments and analysis justifies the effectiveness of the proposed AMDReg for mitigating the effect of data imbalance for generalization to unseen classes in ZS-SBIR.
In this paper, we investigate a new visual restoration task, termed as hallucinating visual instances in total absentia (HVITA). Unlike conventional image inpainting task that works on images with only part of a visual instance missing, HVITA concerns scenarios where an object is completely absent from the scene. This seemingly minor difference in fact makes the HVITA a much challenging task, as the restoration algorithm would have to not only infer the category of the object in total absentia, but also hallucinate an object of which the appearance is consistent with the background. Towards solving HVITA, we propose an end-to-end deep approach that explicitly looks into the global semantics within the image. Specifically, we transform the input image to a semantic graph, wherein each node corresponds to a detected object in the scene. We then adopt a Graph Convolutional Network on top of the scene graph to estimate the category of the missing object in the masked region, and finally introduce a Generative Adversarial Module to carry out the hallucination. Experiments on COCO, Visual Genome and NYU Depth v2 datasets demonstrate that the proposed approach yields truly encouraging and visually plausible results.
Object detection using an oriented bounding box (OBB) can better target rotated objects by reducing the overlap with background areas. Existing OBB approaches are mostly built on horizontal bounding box detectors by introducing an additional angle dimension optimized by a distance loss. However, as the distance loss only minimizes the angle error of the OBB and that it loosely correlates to the IoU, it is insensitive to objects with high aspect ratios. Therefore, a novel loss, Pixels-IoU (PIoU) Loss, is formulated to exploit both the angle and IoU for accurate OBB regression. The PIoU loss is derived from IoU metric with a pixel-wise form, which is simple and suitable for both horizontal and oriented bounding box. To demonstrate its effectiveness, we evaluate the PIoU loss on both anchor-based and anchor-free frameworks. The experimental results show that PIoU loss can dramatically improve the performance of OBB detectors, particularly on objects with high aspect ratios and complex backgrounds. Besides, previous evaluation datasets did not include scenarios where the objects have high aspect ratios, hence a new dataset, Retail50K, is introduced to encourage the community to adapt OBB detectors for more complex environments.
Great progress has been made in 3D body pose and shape estimation from single photos. Yet, state-of-the-art results still suffer from errors due to challenging body poses, modeling clothing, and self occlusions. The domain of basketball games is particularly challenging, due to all of these factors. In this paper, we introduce a new approach for reconstruction of basketball players, that outperforms the state-of-the-art. Key to our approach is new approach for creating poseable, skinned models of NBA players, and a large database of meshes (derived from the NBA2K19 video game), that we are releasing to the research community. Based on these models, we introduce a new method that takes as input a single photo of a clothed player performing any basketball pose and outputs a high resolution mesh and pose of that player. We compare to state of the art methods for shape generation and show significant improvement in the results.
We address the problem of anomaly detection in videos. The goal is to identify unusual behaviours automatically by learning exclusively from normal videos. Most existing approaches are usually data-hungry and have limited generalization abilities. They usually need to be trained on a large number of videos from a target scene to achieve good results in that scene. In this paper, we propose a novel few-shot scene-adaptive anomaly detection problem to address the limitations of previous approaches. Our goal is to learn to detect anomalies in a previously unseen scene with only a few frames. A reliable solution for this new problem will have huge potential in real-world applications since it is expensive to collect a massive amount of data for each target scene. We propose a meta-learning based approach for solving this new problem; extensive experimental results demonstrate the effectiveness of our proposed method.
Humans can robustly recognize and localize objects by integrating visual and auditory cues. While machines are able to do the same now with images, less work has been done with sounds. This work develops an approach for dense semantic labelling of sound-making objects, purely based on binaural sounds. We propose a novel sensor setup and record a new audio-visual dataset of street scenes with eight professional binaural microphones and a $360^{
Semi-Supervised Learning (SSL) based on Convolutional Neural Networks (CNNs) have recently been proven as powerful tools for standard tasks such as image classification when there is not a sufficient amount of labeled data available during the training. In this work, we consider the general setting of the SSL problem for image classification, where the labeled and unlabeled data come from the same underlying distribution. We propose a new SSL method that adopts a hierarchical Optimal Transport (OT) technique to find a mapping from empirical unlabeled measures to corresponding labeled measures by leveraging the minimum amount of transportation cost in the label space. Based on this mapping, pseudo-labels for the unlabeled data are inferred, which are then used along with the labeled data for training the CNN. We evaluated and compared our method with state-of-the-art SSL approaches on standard datasets to demonstrate the superiority of our SSL method.
In this paper, we present a spatial rectifier to estimate surface normals of tilted images. Tilted images are of particular interest as more visual data are captured by arbitrarily oriented sensors such as body-/robot-mounted cameras. Existing approaches exhibit bounded performance on predicting surface normals because they were trained using gravity-aligned images. Our two main hypotheses are: (1) visual scene layout is indicative of the gravity direction; and (2) not all surfaces are equally represented by a learned estimator due to the structured distribution of the training data, i.e., there exists a transformation for each tilted image that is more responsive to the learned estimator than others. We design a spatial rectifier that is learned to transform the surface normal distribution of a tilted image to the rectified one that matches the gravity-aligned training data distribution. Along with the spatial rectifier, we propose a novel truncated angular loss that offers a stronger gradient at small angular errors and robustness to outliers. The resulting estimator outperforms the state-of-the-art methods including data augmentation baselines not only on ScanNet and NYUv2 but also on a new dataset called Tilt-RGBD that includes considerable roll and pitch camera motion.
AutoAugment has been a powerful algorithm that improves the accuracy of many vision tasks, yet it is sensitive to the operator space as well as hyper-parameters, and an improper setting may degenerate network optimization. This paper delves deep into the working mechanism, and reveals that AutoAugment may remove part of discriminative information from the training image and so insisting on the ground-truth label is no longer the best option. To relieve the inaccuracy of supervision, we make use of knowledge distillation that refers to the output of a teacher model to guide network training. Experiments are performed in standard image classification benchmarks, and demonstrate the effectiveness of our approach in suppressing noise of data augmentation and stabilizing training. Upon the cooperation of knowledge distillation and AutoAugment, we claim the new state-of-the-art on ImageNet classification with a top-1 accuracy of 85.7%.
Open set recognition is an emerging research area that aims to simultaneously classify samples from predefined classes and identify the rest as 'unknown'. In this process, one of the key challenges is to reduce the risk of generalizing the inherent characteristics of numerous unknown samples learned from a small amount of known data. In this paper, we propose a new concept, Reciprocal Point, which is the potential representation of the extra-class space corresponding to each known category. The sample can be classified to known or unknown by the otherness with reciprocal points. To tackle the open set problem, we offer a novel open space risk regularization term. Based on the bounded space constructed by reciprocal points, the risk of unknown is reduced through multi-category interaction. The novel learning framework called Reciprocal Point Learning (RPL), which can indirectly introduce the unknown information into the learner with only known classes, so as to learn more compact and discriminative representations. Moreover, we further construct a new large-scale challenging aircraft dataset for open set recognition: Aircraft 300 (Air-300). Extensive experiments on multiple benchmark datasets indicate that our framework is significantly superior to other existing approaches and achieves state-of-the-art performance on standard open set benchmarks.
Existing interactive object segmentation methods mainly take spatial interactions such as bounding boxes or clicks as input. However, these interactions do not contain information about explicit attributes of the target-of-interest and thus cannot quickly specify what the selected object exactly is, especially when there are diverse scales of candidate objects or the target-of-interest contains multiple objects. Therefore, excessive user interactions are often required to reach desirable results. On the other hand, in existing approaches attribute information of objects is often not well utilized in interactive segmentation. We propose to employ phrase expressions as another interaction input to infer the attributes of target object. In this way, we can 1) leverage spatial clicks to locate the target object and 2) utilize semantic phrases to qualify the attributes of the target object. Specifically, the phrase expressions focus on ``what"" the target object is and the spatial clicks are in charge of ``where"" the target object is, which together help to accurately segment the target-of-interest with smaller number of interactions. Moreover, the proposed approach is flexible in terms of interaction modes and can efficiently handle complex scenarios by leveraging the strengths of each type of input. Our multi-modal phrase+click approach achieves new state-of-the-art performance on interactive segmentation. To the best of our knowledge, this is the first work to leverage both clicks and phrases for interactive segmentation.
This paper introduces data augmentation for point clouds by interpolation between examples. Data augmentation by interpolation has shown to be a simple and effective approach in the image domain. Such a mixup is however not directly transferable to point clouds, as we do not have a one-to-one correspondence between the points of two different objects. In this paper, we define data augmentation between point clouds as a shortest path linear interpolation. To that end, we introduce PointMixup, an interpolation method that generates new examples through an optimal assignment of the path function between two point clouds. We prove that our PointMixup finds the shortest path between two point clouds and that the interpolation is assignment invariant and linear. With the definition of interpolation, PointMixup allows to introduce strong interpolation-based regularizers such as mixup and manifold mixup to the point cloud domain. Experimentally, we show the potential of PointMixup for point cloud classification, especially when examples are scarce, as well as increased robustness to noise and geometric transformations to points. The code for PointMixup and the experimental details are publicly available.