CVPR 2018 论文列表

We propose a method for unsupervised video object segmentation by transferring the knowledge encapsulated in image-based instance embedding networks. The instance embedding network produces an embedding vector for each pixel that enables identifying all pixels belonging to the same object. Though trained on static images, the instance embeddings are stable over consecutive video frames, which allows us to link objects together over time. Thus, we adapt the instance networks trained on static images to video object segmentation and incorporate the embeddings with objectness and optical flow features, without model retraining or online fine-tuning. The proposed method outperforms state-of-the-art unsupervised segmentation methods in the DAVIS dataset and the FBMS dataset.

This paper addresses a new problem of inferring shared attention in third-person social scene videos. Shared attention is a phenomenon that two or more individuals simultaneously look at a common target in social scenes. Perceiving and identifying shared attention in videos plays crucial roles in social activities and social scene understanding. We propose a spatial-temporal neural network to detect shared attention intervals in videos and predict shared attention locations in frames. In each video frame, human gaze directions and potential target boxes are two key features for spatially detecting shared attention in the social scene. In temporal domain, a convolutional Long Short- Term Memory network utilizes the temporal continuity and transition constraints to optimize the predicted shared attention heatmap. We collect a new dataset VideoCoAtt from public TV show videos, containing 380 complex video sequences with more than 492,000 frames that include diverse social scenes for shared attention study. Experiments on this dataset show that our model can effectively infer shared attention in videos. We also empirically verify the effectiveness of different components in our model.

We propose a novel hierarchical approach for text-to-image synthesis by inferring semantic layout. Instead of learning a direct mapping from text to image, our algorithm decomposes the generation process into multiple steps, in which it first constructs a semantic layout from the text by the layout generator and converts the layout to an image by the image generator. The proposed layout generator progressively constructs a semantic layout in a coarse-to-fine manner by generating object bounding boxes and refining each box by estimating object shapes inside the box. The image generator synthesizes an image conditioned on the inferred semantic layout, which provides a useful semantic structure of an image matching with the text description. Our model not only generates semantically more meaningful images, but also allows automatic annotation of generated images and user-controlled generation process by modifying the generated scene layout. We demonstrate the capability of the proposed model on challenging MS-COCO dataset and show that the model can substantially improve the image quality, interpretability of output and semantic alignment to input text over existing approaches.

We present a method for inferring a 4D light field of a hidden scene from 2D shadows cast by a known occluder on a diffuse wall. We do this by determining how light naturally reflected off surfaces in the hidden scene interacts with the occluder. By modeling the light transport as a linear system, and incorporating prior knowledge about light field structures, we can invert the system to recover the hidden scene. We demonstrate results of our inference method across simulations and experiments with different types of occluders. For instance, using the shadow cast by a real house plant, we are able to recover low resolution light fields with different levels of texture and parallax complexity. We provide two experimental results: a human subject and two planar elements at different depths.

Higher Order MRF-MAP formulation has been a popular technique for solving many problems in computer vision. Inference in a general MRF-MAP problem is NP Hard, but can be performed in polynomial time for the special case when potential functions are submodular. Two popular combinatorial approaches for solving such formulations are flow based and polyhedral approaches. Flow based approaches work well with small cliques and in that mode can handle problems with millions of variables. Polyhedral approaches can handle large cliques but in small numbers. We show in this paper that the variables in these seemingly disparate techniques can be mapped to each other. This allows us to combine the two styles in a joint framework exploiting the strength of both of them. Using the proposed joint framework, we are able to perform tractable inference in MRF-MAP problems with millions of variables and a mix of small and large cliques, a formulation which can not be solved by either of the two styles individually. We show applicability of this hybrid framework on object segmentation problem as an example of a situation where quality of results is significantly better than systems which are based only on the use of small or large cliques.

We propose a novel approach to estimate the three degrees of freedom (DoF) drift-free rotational motion of an RGB-D camera from only a single line and plane in the Manhattan world (MW). Previous approaches exploit the surface normal vectors and vanishing points to achieve accurate 3-DoF rotation estimation. However, they require multiple orthogonal planes or many consistent lines to be visible throughout the entire rotation estimation process; otherwise, these approaches fail. To overcome these limitations, we present a new method that estimates absolute camera orientation from only a single line and a single plane in RANSAC, which corresponds to the theoretical minimal sampling for 3-DoF rotation estimation. Once we find an initial rotation estimate, we refine the camera orientation by minimizing the average orthogonal distance from the endpoints of the lines parallel to the MW axes. We demonstrate the effectiveness of the proposed algorithm through an extensive evaluation on a variety of RGB-D datasets and compare with other state-of-the-art methods.

Recurrent neural networks (RNNs) have been widely used for processing sequential data. However, RNNs are commonly difficult to train due to the well-known gradient vanishing and exploding problems and hard to learn long-term patterns. Long short-term memory (LSTM) and gated recurrent unit (GRU) were developed to address these problems, but the use of hyperbolic tangent and the sigmoid action functions results in gradient decay over layers. Consequently, construction of an efficiently trainable deep network is challenging. In addition, all the neurons in an RNN layer are entangled together and their behaviour is hard to interpret. To address these problems, a new type of RNN, referred to as independently recurrent neural network (IndRNN), is proposed in this paper, where neurons in the same layer are independent of each other and they are connected across layers. We have shown that an IndRNN can be easily regulated to prevent the gradient exploding and vanishing problems while allowing the network to learn long-term dependencies. Moreover, an IndRNN can work with non-saturated activation functions such as relu (rectified linear unit) and be still trained robustly. Multiple IndRNNs can be stacked to construct a network that is deeper than the existing RNNs. Experimental results have shown that the proposed IndRNN is able to process very long sequences (over 5000 time steps), can be used to construct very deep networks (21 layers used in the experiment) and still be trained robustly. Better performances have been achieved on various tasks by using IndRNNs compared with the traditional RNN and LSTM.

We seek to predict the 6 degree-of-freedom (6DoF) pose of a query photograph with respect to a large indoor 3D map. The contributions of this work are three-fold. First, we develop a new large-scale visual localization method targeted for indoor environments. The method proceeds along three steps: (i) efficient retrieval of candidate poses that ensures scalability to large-scale environments, (ii) pose estimation using dense matching rather than local features to deal with textureless indoor scenes, and (iii) pose verification by virtual view synthesis to cope with significant changes in viewpoint, scene layout, and occluders. Second, we collect a new dataset with reference 6DoF poses for large-scale indoor localization. Query photographs are captured by mobile phones at a different time than the reference 3D map, thus presenting a realistic indoor localization scenario. Third, we demonstrate that our method significantly outperforms current state-of-the-art indoor localization approaches on this new challenging data.

In this work we present In-Place Activated Batch Normalization (InPlace-ABN) -- a novel approach to drastically reduce the training memory footprint of modern deep neural networks in a computationally efficient way. Our solution substitutes the conventionally used succession of BatchNorm + Activation layers with a single plugin layer, hence avoiding invasive framework surgery while providing straightforward applicability for existing deep learning frameworks. We obtain memory savings of up to 50% by dropping intermediate results and by recovering required information during the backward pass through the inversion of stored forward results, with only minor increase (0.8-2%) in computation time. Also, we demonstrate how frequently used checkpointing approaches can be made computationally as efficient as InPlace-ABN. In our experiments on image classification, we demonstrate on-par results on ImageNet-1k with state-of-the-art approaches. On the memory-demanding task of semantic segmentation, we report competitive results for COCO-Stuff and set new state-of-the-art results for Cityscapes and Mapillary Vistas. Code can be found at https://github.com/mapillary/inplace_abn.

We propose methods of addressing two critical issues of pedestrian detection: (i) occlusion of target objects as false negative failure, and (ii) confusion with hard negative examples like vertical structures as false positive failure. Our solutions to these two problems are general and flexible enough to be applicable to any single-stage detection models. We implement our methods into four state-of-the-art single-stage models, including SqueezeDet+, YOLOv2, SSD, and DSSD. We empirically validate that our approach indeed improves the performance of those four models on Caltech pedestrian and CityPersons dataset. Moreover, in some heavy occlusion settings, our approach achieves the best reported performance. Specifically, our two solutions are as follows. For better occlusion handling, we update the output tensors of single-stage models so that they include the prediction of part confidence scores, from which we compute a final occlusion-aware detection score. For reducing confusion with hard negative examples, we introduce average grid classifiers as post-refinement classifiers, trainable in an end-to-end fashion with little memory and time overhead (e.g. increase of 1--5 MB in memory and 1--2 ms in inference time).

We demonstrate that many detection methods are designed to identify only a sufficently accurate bounding box, rather than the best available one. To address this issue we propose a simple and fast modification to the existing methods called Fitness NMS. This method is tested with the DeNet model and obtains a significantly improved MAP at greater localization accuracies without a loss in evaluation rate, and can be used in conjunction with Soft NMS for additional improvements. Next we derive a novel bounding box regression loss based on a set of IoU upper bounds that better matches the goal of IoU maximization while still providing good convergence properties. Following these novelties we investigate RoI clustering schemes for improving evaluation rates for the DeNet wide model variants and provide an analysis of localization performance at various input image dimensions. We obtain a MAP of 33.6%@79Hz and 41.8%@5Hz for MSCOCO and a Titan X (Maxwell).

We present two techniques to improve landmark localization in images from partially annotated datasets. Our primary goal is to leverage the common situation where precise landmark locations are only provided for a small data subset, but where class labels for classification or regression tasks related to the landmarks are more abundantly available. First, we propose the framework of sequential multitasking and explore it here through an architecture for landmark localization where training with class labels acts as an auxiliary signal to guide the landmark localization on unlabeled data. A key aspect of our approach is that errors can be backpropagated through a complete landmark localization model. Second, we propose and explore an unsupervised learning technique for landmark localization based on having a model predict equivariant landmarks with respect to transformations applied to the image. We show that these techniques, improve landmark prediction considerably and can learn effective detectors even when only a small fraction of the dataset has landmark labels. We present results on two toy datasets and four real datasets, with hands and faces, and report new state-of-the-art on two datasets in the wild, e.g. with only 5% of labeled images we outperform previous state-of-the-art trained on the AFLW dataset.

One of the key operations performed on a digital camera is to map the sensor-specific color space to a standard perceptual color space. This procedure involves the application of a white-balance correction followed by a color space transform. The current approach for this colorimetric mapping is based on an interpolation of pre-calibrated color space transforms computed for two fixed illuminations (i.e., two white-balance settings). Images captured under different illuminations are subject to less color accuracy due to the use of this interpolation process. In this paper, we discuss the limitations of the current colorimetric mapping approach and propose two methods that are able to improve color accuracy. We evaluate our approach on seven different cameras and show improvements of up to 30% (DSLR cameras) and 59% (mobile phone cameras) in terms of color reproduction error.

We develop a set of methods to improve on the results of self-supervised learning using context. We start with a baseline of patch based arrangement context learning and go from there. Our methods address some overt problems such as chromatic aberration as well as other potential problems such as spatial skew and mid-level feature neglect. We prevent problems with testing generalization on common self-supervised benchmark tests by using different datasets during our development. The results of our methods combined yield top scores on all standard self-supervised benchmarks, including classification and detection on PASCAL VOC 2007, segmentation on PASCAL VOC 2012, and "linear tests" on the ImageNet and CSAIL Places datasets. We obtain an improvement over our baseline method of between 4.0 to 7.1 percentage points on transfer learning classification tests. We also show results on different standard network architectures to demonstrate generalization as well as portability. All data, models and programs are available at: https://gdo-datasci. llnl.gov/selfsupervised/.

We propose a method for lossy image compression based on recurrent, convolutional neural networks that outper- forms BPG (4:2:0), WebP, JPEG2000, and JPEG as mea- sured by MS-SSIM. We introduce three improvements over previous research that lead to this state-of-the-art result us- ing a single model. First, we modify the recurrent architec- ture to improve spatial diffusion, which allows the network to more effectively capture and propagate image informa- tion through the network’s hidden state. Second, in addition to lossless entropy coding, we use a spatially adaptive bit allocation algorithm to more efficiently use the limited num- ber of bits to encode visually complex image regions. Fi- nally, we show that training with a pixel-wise loss weighted by SSIM increases reconstruction quality according to sev- eral metrics. We evaluate our method on the Kodak and Tecnick image sets and compare against standard codecs as well as recently published methods based on deep neural networks.

A key solution to visual question answering (VQA) exists in how to fuse visual and language features extracted from an input image and question. We show that an attention mechanism that enables dense, bi-directional interactions between the two modalities contributes to boost accuracy of prediction of answers. Specifically, we present a simple architecture that is fully symmetric between visual and language representations, in which each question word attends on image regions and each image region attends on question words. It can be stacked to form a hierarchy for multi-step interactions between an image-question pair. We show through experiments that the proposed architecture achieves a new state-of-the-art on VQA and VQA 2.0 despite its small size. We also present qualitative evaluation, demonstrating how the proposed attention mechanism can generate reasonable attention maps on images and questions, which leads to the correct answer prediction.

This paper proposes an importance weighted adversarial nets-based method for unsupervised domain adaptation, specific for partial domain adaptation where the target domain has less number of classes compared to the source domain. Previous domain adaptation methods generally assume the identical label spaces, such that reducing the distribution divergence leads to feasible knowledge transfer. However, such an assumption is no longer valid in a more realistic scenario that requires adaptation from a larger and more diverse source domain to a smaller target domain with less number of classes. This paper extends the adversarial nets-based domain adaptation and proposes a novel adversarial nets-based partial domain adaptation method to identify the source samples that are potentially from the outlier classes and, at the same time, reduce the shift of shared classes between domains.

Person re-identification (re-ID) models trained on one domain often fail to generalize well to another. In our attempt, we present a ``learning via translation'' framework. In the baseline, we translate the labeled images from source to target domain in an unsupervised manner. We then train re-ID models with the translated images by supervised methods. Yet, being an essential part of this framework, unsupervised image-image translation suffers from the information loss of source-domain labels during translation. Our motivation is two-fold. First, for each image, the discriminative cues contained in its ID label should be maintained after translation. Second, given the fact that two domains have entirely different persons, a translated image should be dissimilar to any of the target IDs. To this end, we propose to preserve two types of unsupervised similarities, 1) self-similarity of an image before and after translation, and 2) domain-dissimilarity of a translated source image and a target image. Both constraints are implemented in the similarity preserving generative adversarial network (SPGAN) which consists of an Siamese network and a CycleGAN. Through domain adaptation experiment, we show that images generated by SPGAN are more suitable for domain adaptation and yield consistent and competitive re-ID accuracy on two large-scale datasets.

We propose a general framework for unsupervised domain adaptation, which allows deep neural networks trained on a source domain to be tested on a different target domain without requiring any training annotations in the target domain. This is achieved by adding extra networks and losses that help regularize the features extracted by the backbone encoder network. To this end we propose the novel use of the recently proposed unpaired image-to-image translation framework to constrain the features extracted by the encoder network. Specifically, we require that the features extracted are able to reconstruct the images in both domains. In addition we require that the distribution of features extracted from images in the two domains are indistinguishable. Many recent works can be seen as specific cases of our general framework. We apply our method for domain adaptation between MNIST, USPS, and SVHN datasets, and Amazon, Webcam and DSLR Office datasets in classification tasks, and also between GTA5 and Cityscapes datasets for a segmentation task. We demonstrate state of the art performance on each of these datasets.