CVPR 2017 Papers

We focus on the non-Lambertian object-level intrinsic problem of recovering diffuse albedo, shading, and specular highlights from a single image of an object. Based on existing 3D models in the ShapeNet database, a large-scale object intrinsics database is rendered with HDR environment maps. Millions of synthetic images of objects and their corresponding albedo, shading, and specular ground-truth images are used to train an encoder-decoder CNN, which can decompose an image into the product of albedo and shading components along with an additive specular component. Our CNN delivers accurate and sharp results in this classical inverse problem of computer vision. Evaluated on our realistically synthetic dataset, our method consistently outperforms the state-of-the-art by a large margin. We train and test our CNN across different object categories. Perhaps surprising especially from the CNN classification perspective, our intrinsics CNN generalizes very well across categories. Our analysis shows that feature learning at the encoder stage is more crucial for developing a universal representation across categories. We apply our model to real images and videos from Internet, and observe robust and realistic intrinsics results. Quality non-Lambertian intrinsics could open up many interesting applications such as realistic product search based on material properties and image-based albedo / specular editing.

In this paper we propose a unified framework to address multiple realistic image retrieval tasks concerning both category and attributes. Considering the scale of modern datasets, hashing is favorable for its low complexity. However, most existing hashing methods are designed to preserve one single kind of similarity, thus incapable of dealing with the different tasks simultaneously. To overcome this limitation, we propose a new hashing method, named Dual Purpose Hashing (DPH), which jointly preserves the category and attribute similarities by exploiting the convolutional networks (CNN) to hierarchically capture the correlations between category and attributes. Since images with both category and attribute labels are scarce, our method is designed to take the abundant partially labelled images on the Internet as training inputs. With such a framework, the binary codes of new-coming images can be readily obtained by quantizing the network outputs of a binary-like layer, and the attributes can be recovered from the codes easily. Experiments on two large-scale datasets show that our dual purpose hash codes can achieve comparable or even better performance than those state-of-the-art methods specifically designed for each individual retrieval task, while being more compact than the compared methods.

The problem of determining whether an object is in motion, irrespective of camera motion, is far from being solved. We address this challenging task by learning motion patterns in videos. The core of our approach is a fully convolutional network, which is learned entirely from synthetic video sequences, and their ground-truth optical flow and motion segmentation. This encoder-decoder style architecture first learns a coarse representation of the optical flow field features, and then refines it iteratively to produce motion labels at the original high-resolution. We further improve this labeling with an objectness map and a conditional random field, to account for errors in optical flow, and also to focus on moving "things" rather than "stuff". The output label of each pixel denotes whether it has undergone independent motion, i.e., irrespective of camera motion. We demonstrate the benefits of this learning framework on the moving object segmentation task, where the goal is to segment all objects in motion. Our approach outperforms the top method on the recently released DAVIS benchmark dataset, comprising real-world sequences, by 5.6%. We also evaluate on the Berkeley motion segmentation database, achieving state-of-the-art results.

In this paper, we propose a fully convolutional network for iterative non-blind deconvolution. We decompose the non-blind deconvolution problem into image denoising and image deconvolution. We train a FCNN to remove noise in the gradient domain and use the learned gradients to guide the image deconvolution step. In contrast to the existing deep neural network based methods, we iteratively deconvolve the blurred images in a multi-stage framework. The proposed method is able to learn an adaptive image prior, which keeps both local (details) and global (structures) information. Both quantitative and qualitative evaluations on the benchmark datasets demonstrate that the proposed method performs favorably against state-of-the-art algorithms in terms of quality and speed.

Estimating human pose, shape, and motion from images and video are fundamental challenges with many applications. Recent advances in 2D human pose estimation use large amounts of manually-labeled training data for learning convolutional neural networks (CNNs). Such data is time consuming to acquire and difficult to extend. Moreover, manual labeling of 3D pose, depth and motion is impractical. In this work we present SURREAL: a new large-scale dataset with synthetically-generated but realistic images of people rendered from 3D sequences of human motion capture data. We generate more than 6 million frames together with ground truth pose, depth maps, and segmentation masks. We show that CNNs trained on our synthetic dataset allow for accurate human depth estimation and human part segmentation in real RGB images. Our results and the new datast open up new possibilities for advancing person analysis using chap and large-scale synthetic data.

With recent progress in graphics, it has become more tractable to train models on synthetic images, potentially avoiding the need for expensive annotations. However, learning from synthetic images may not achieve the desired performance due to a gap between synthetic and real image distributions. To reduce this gap, we propose Simulated+Unsupervised (S+U) learning, where the task is to learn a model to improve the realism of a simulator's output using unlabeled real data, while preserving the annotation information from the simulator. We develop a method for S+U learning that uses an adversarial network similar to Generative Adversarial Networks (GANs), but with synthetic images as inputs instead of random vectors. We make several key modifications to the standard GAN algorithm to preserve annotations, avoid artifacts, and stabilize training: (i) a 'self-regularization' term, (ii) a local adversarial loss, and (iii) updating the discriminator using a history of refined images. We show that this enables generation of highly realistic images, which we demonstrate both qualitatively and with a user study. We quantitatively evaluate the generated images by training models for gaze estimation and hand pose estimation. We show a significant improvement over using synthetic images, and achieve state-of-the-art results on the MPIIGaze dataset without any labeled real data.

We present an approach to effectively use millions of images with noisy annotations in conjunction with a small subset of cleanly-annotated images to learn powerful image representations. One common approach to combine clean and noisy data is to first pre-train a network using the large noisy dataset and then fine-tune with the clean dataset. We show this approach does not fully leverage the information contained in the clean set. Thus, we demonstrate how to use the clean annotations to reduce the noise in the large dataset before fine-tuning the network using both the clean set and the full set with reduced noise. The approach comprises a multi-task network that jointly learns to clean noisy annotations and to accurately classify images. We evaluate our approach on the recently released Open Images dataset, containing 9 million images, multiple annotations per image and over 6000 unique classes. For the small clean set of annotations we use a quarter of the validation set with 40k images. Our results demonstrate that the proposed approach clearly outperforms direct fine-tuning across all major categories of classes in the Open Image dataset. Further, our approach is particularly effective for a large number of classes with wide range of noise in annotations (20-80% false positive annotations).

This paper presents a novel yet intuitive approach to unsupervised feature learning. Inspired by the human visual system, we explore whether low-level motion-based grouping cues can be used to learn an effective visual representation. Specifically, we use unsupervised motion-based segmentation on videos to obtain segments, which we use as 'pseudo ground truth' to train a convolutional network to segment objects from a single frame. Given the extensive evidence that motion plays a key role in the development of the human visual system, we hope that this straightforward approach to unsupervised learning will be more effective than cleverly designed 'pretext' tasks studied in the literature. Indeed, our extensive experiments show that this is the case. When used for transfer learning on object detection, our representation significantly outperforms previous unsupervised approaches across multiple settings, especially when training data for the target task is scarce.

The depth images acquired by consumer depth sensors (e.g., Kinect and ToF) usually are of low resolution and insufficient quality. One natural solution is to incorporate with high resolution RGB camera for exploiting their statistical correlation. However, most existing methods are intuitive and limited in characterizing the complex and dynamic dependency between intensity and depth images. To address these limitations, we propose a weighted analysis representation model for guided depth image enhancement, which advances the conventional methods in two aspects: (i) task driven learning and (ii) dynamic guidance. First, we generalize the analysis representation model by including a guided weight function for dependency modeling. And the task-driven learning formulation is introduced to obtain the optimized guidance tailored to specific enhancement task. Second, the depth image is gradually enhanced along with the iterations, and thus the guidance should also be dynamically adjusted to account for the updating of depth image. To this end, stage-wise parameters are learned for dynamic guidance. Experiments on guided depth image upsampling and noisy depth image restoration validate the effectiveness of our method.

Colorization is an ambiguous problem, with multiple viable colorizations for a single grey-level image. However, previous methods only produce the single most probable colorization. Our goal is to model the diversity intrinsic to the problem of colorization and produce multiple colorizations that display long-scale spatial co-ordination. We learn a low dimensional embedding of color fields using a variational autoencoder (VAE). We construct loss terms for the VAE decoder that avoid blurry outputs and take into account the uneven distribution of pixel colors. Finally, we build a conditional model for the multi-modal distribution between grey-level image and the color field embeddings. Samples from this conditional model result in diverse colorization. We demonstrate that our method obtains better diverse colorizations than a standard conditional variational autoencoder (CVAE) model, as well as a recently proposed conditional generative adversarial network (cGAN).

Robust covariant local feature detectors are important for detecting local features that are (1) discriminative of the image content and (2) can be repeatably detected at consistent locations when the image undergoes diverse transformations. Such detectors are critical for applications such as image search and scene reconstruction. Many learning-based local feature detectors address one of these two problems while overlooking the other. In this work, we propose a novel learning-based method to simultaneously address both issues. Specifically, we extend the covariant constraint proposed by Lenc and Vedaldi by defining the concepts of "standard patch" and "canonical feature" and leverage these to train a novel robust covariant detector. We show that the introduction of these concepts greatly simplifies the learning stage of the covariant detector, and also makes the detector much more robust. Extensive experiments show that our method outperforms previous hand-crafted and learning-based detectors by large margins in terms of repeatability.

To predict a set of diverse and informative proposals with enriched representations, this paper introduces a differentiable Determinantal Point Process (DPP) layer that is able to augment the object detection architectures. Most modern object detection architectures, such as Faster R-CNN, learn to localize objects by minimizing deviations from the ground truth, but ignore correlation between multiple proposals and object categories. Non-Maximum Suppression (NMS) as a widely used proposal pruning scheme ignores label- and instance-level relations between object candidates resulting in multi-labeled detections. In the multi-class case, NMS selects boxes with the largest prediction scores ignoring the semantic relation between categories of potential election. In contrast, our trainable DPP layer, allowing for Learning Detection with Diverse Proposals (LDDP), considers both label-level contextual information and spatial layout relationships between proposals without increasing the number of parameters of the network, and thus improves location and category specifications of final detected bounding boxes substantially during both training and inference schemes. Furthermore, we show that LDDP keeps it superiority over Faster R-CNN even if the number of proposals generated by LDPP is only 30% as many as those for Faster R-CNN.

Reconstructing the detailed geometric structure of a face from a given image is a key to many computer vision and graphics applications, such as motion capture and reenactment. The reconstruction task is challenging as human faces vary extensively when considering expressions, poses, textures, and intrinsic geometries. While many approaches tackle this complexity by using additional data to reconstruct the face of a single subject, extracting facial surface from a single image remains a difficult problem. As a result, single-image based methods can usually provide only a rough estimate of the facial geometry. In contrast, we propose to leverage the power of convolutional neural networks to produce a highly detailed face reconstruction from a single image. For this purpose, we introduce an end-to-end CNN framework which derives the shape in a coarse-to-fine fashion. The proposed architecture is composed of two main blocks, a network that recovers the coarse facial geometry (CoarseNet), followed by a CNN that refines the facial features of that geometry (FineNet). The proposed networks are connected by a novel layer which renders a depth image given a mesh in 3D. Unlike object recognition and detection problems, there are no suitable datasets for training CNNs to perform face geometry reconstruction. Therefore, our training regime begins with a supervised phase, based on synthetic images, followed by an unsupervised phase that uses only unconstrained facial images. The accuracy and robustness of the proposed model is demonstrated by both qualitative and quantitative evaluation tests.

Image-set classification has recently generated great popularity due to its widespread applications in computer vision. The great challenges arise from effectively and efficiently measuring the similarity between image sets with high inter-class ambiguity and huge intra-class variability. In this paper, we propose deep match kernels (DMK) to directly measure the similarity between image sets in the match kernel framework. Specifically, we build deep local match kernels between images upon arc-cosine kernels, which can faithfully characterize the similarity between images by mimicking deep neural networks; we introduce anchors to aggregate those deep local match kernels into a global match kernel between image sets, which is learned in a supervised way by kernel alignment and therefore more discriminative. The DMK provides the first match kernel framework for image-set classification, which removes specific assumptions usually required in previous approaches and is computationally more efficient. We conduct extensive experiments on four datasets for three diverse image-set classification tasks. The DMK achieves high performance and consistently surpasses state-of-the-art methods, showing its great effectiveness for image-set classification.

Person Re-identification (ReID) is to identify the same person across different cameras. It is a challenging task due to the large variations in person pose, occlusion, background clutter, etc. How to extract powerful features is a fundamental problem in ReID and is still an open problem today. In this paper, we design a Multi-Scale Context-Aware Network (MSCAN) to learn powerful features over full body and body parts, which can well capture the local context knowledge by stacking multi-scale convolutions in each layer. Moreover, instead of using predefined rigid parts, we propose to learn and localize deformable pedestrian parts using Spatial Transformer Networks (STN) with novel spatial constraints. The learned body parts can release some difficulties, e.g. pose variations and background clutters, in part-based representation. Finally, we integrate the representation learning processes of full body and body parts into a unified framework for person ReID through multi-class person identification tasks. Extensive evaluations on current challenging large-scale person ReID datasets, including the image-based Market1501, CUHK03 and sequence-based MARS datasets, show that the proposed method achieves the state-of-the-art results.

Model-based optimization methods and discriminative learning methods have been the two dominant strategies for solving various inverse problems in low-level vision. Typically, those two kinds of methods have their respective merits and drawbacks, e.g., model-based optimization methods are flexible for handling different inverse problems but are usually time-consuming with sophisticated priors for the purpose of good performance; in the meanwhile, discriminative learning methods have fast testing speed but their application range is greatly restricted by the specialized task. Recent works have revealed that, with the aid of variable splitting techniques, denoiser prior can be plugged in as a modular part of model-based optimization methods to solve other inverse problems (e.g., deblurring). Such an integration induces considerable advantage when the denoiser is obtained via discriminative learning. However, the study of integration with fast discriminative denoiser prior is still lacking. To this end, this paper aims to train a set of fast and effective CNN (convolutional neural network) denoisers and integrate them into model-based optimization method to solve other inverse problems. Experimental results demonstrate that the learned set of denoisers can not only achieve promising Gaussian denoising results but also can be used as prior to deliver good performance for various low-level vision applications.

In this paper, we propose an unsupervised feature learning method called deep binary descriptor with multi-quantization (DBD-MQ) for visual matching. Existing learning-based binary descriptors such as compact binary face descriptor (CBFD) and DeepBit utilize the rigid sign function for binarization despite of data distributions, thereby suffering from severe quantization loss. In order to address the limitation, our DBD-MQ considers the binarization as a multi-quantization task. Specifically, we apply a K-AutoEncoders (KAEs) network to jointly learn the parameters and the binarization functions under a deep learning framework, so that discriminative binary descriptors can be obtained with a fine-grained multi-quantization. Extensive experimental results on different visual analysis including patch retrieval, image matching and image retrieval show that our DBD-MQ outperforms most existing binary feature descriptors.

In this paper, we introduce Recipe1M, a new large-scale, structured corpus of over 1m cooking recipes and 800k food images. As the largest publicly available collection of recipe data, Recipe1M affords the ability to train high-capacity models on aligned, multi-modal data. Accordingly, we train a neural network to find a joint embedding of recipes and images that yields impressive results on an image-recipe retrieval task. Additionally, we demonstrate that regularization via the addition of a high-level, semantic classification objective improves performance to rival that of humans and enables semantic vector arithmetic. We postulate that these embeddings will provide a basis for further exploration of the Recipe1M dataset and food and cooking in general.

This paper presents a novel method for detecting pedestrians under adverse illumination conditions. Our approach relies on a novel cross-modality learning framework and it is based on two main phases. First, given a multimodal dataset, a deep convolutional network is employed to learn a non-linear mapping, modeling the relations between RGB and thermal data. Then, the learned feature representations are transferred to a second deep network, which receives as input an RGB image and outputs the detection results. In this way, features which are both discriminative and robust to bad illumination conditions are learned. Importantly, at test time, only the second pipeline is considered and no thermal data are required. Our extensive evaluation demonstrates that the proposed approach outperforms the state-of-the-art on the challenging KAIST multispectral pedestrian dataset and it is competitive with previous methods on the popular Caltech dataset.