ICCV 2017 Papers

Most state-of-the-art motion segmentation algorithms draw their potential from modeling motion differences of local entities such as point trajectories in terms of pairwise potentials in graphical models. Inference in instances of minimum cost multicut problems defined on such graphs allows to optimize the number of the resulting segments along with the segment assignment. However, pairwise potentials limit the discriminative power of the employed motion models to translational differences. More complex models such as Euclidean or affine transformations call for higher-order potentials and a tractable inference in the resulting higher-order graphical models. In this paper, we (1) introduce a generalization of the minimum cost lifted multicut problem to hypergraphs, and (2) propose a simple primal feasible heuristic that allows for a reasonably efficient inference in instances of higher-order lifted multicut problem instances defined on point trajectory hypergraphs for motion segmentation. The resulting motion segmentations improve over the state-of-the-art on the FBMS-59 dataset.

The success of fine-grained visual categorization (FGVC) extremely relies on the modeling of appearance and interactions of various semantic parts. This makes FGVC very challenging because: (i) part annotation and detection require expert guidance and are very expensive; (ii) parts are of different sizes; and (iii) the part interactions are complex and of higher-order. To address these issues, we propose an end-to-end framework based on higher-order integration of hierarchical convolutional activations for FGVC. By treating the convolutional activations as local descriptors, hierarchical convolutional activations can serve as a representation of local parts from different scales. A polynomial kernel based predictor is proposed to capture higher-order statistics of convolutional activations for modeling part interaction. To model inter-layer part interactions, we extend polynomial predictor to integrate hierarchical activations via kernel fusion. Our work also provides a new perspective for combining convolutional activations from multiple layers. While hypercolumns simply concatenate maps from different layers, and holistically-nested network uses weighted fusion to combine side-outputs, our approach exploits higher-order intra-layer and inter-layer relations for better integration of hierarchical convolutional features. The proposed framework yields more discriminative representation and achieves competitive results on the widely used FGVC datasets.

Estimating correspondence between two images and extracting the foreground object are two challenges in computer vision. With dual-lens smart phones, such as iPhone 7Plus and Huawei P9, coming into the market, two images of slightly different views provide us new information to unify the two topics. We propose a joint method to tackle them simultaneously via a joint fully connected conditional random field (CRF) framework. The regional correspondence is used to handle textureless regions in matching and make our CRF system computationally efficient. Our method is evaluated over 2,000 new image pairs, and produces promising results on challenging portrait images.

Convolutional sparse coding (CSC) has gained attention for its successful role as a reconstruction and a classification tool in the computer vision and machine learning community. Current CSC methods can only reconstruct single-feature 2D images independently. However, learning multi-dimensional dictionaries and sparse codes for the reconstruction of multi-dimensional data is very important, as it examines correlations among all the data jointly. This provides more capacity for the learned dictionaries to better reconstruct data. In this paper, we propose a generic and novel formulation for the CSC problem that can handle an arbitrary order tensor of data. Backed with experimental results, our proposed formulation can not only tackle applications that are not possible with standard CSC solvers, including colored video reconstruction (5D- tensors), but it also performs favorably in reconstruction with much fewer parameters as compared to naive extensions of standard CSC to multiple features/channels.

We address the problem of dense visual-semantic embedding that maps not only full sentences and whole images but also phrases within sentences and salient regions within images into a multimodal embedding space. As a result, we can produce several region-oriented and expressive phrases rather than just an overview sentence to describe an image. In particular, we present a hierarchical structured recurrent neural network (RNN), namely Hierarchical Multimodal LSTM (HM-LSTM) model. Different from chain structured RNN, our model presents a hierarchical structure so that it can naturally build representations for phrases and image regions, and further exploit their hierarchical relations. Moreover, the fine-grained correspondences between phrases and image regions can be automatically learned and utilized to boost the learning of the dense embedding space. Extensive experiments on several datasets validate the efficacy of our proposed method, which compares favorably with the state-of-the-art methods.

We propose 'Hide-and-Seek', a weakly-supervised framework that aims to improve object localization in images and action localization in videos. Most existing weakly-supervised methods localize only the most discriminative parts of an object rather than all relevant parts, which leads to suboptimal performance. Our key idea is to hide patches in a training image randomly, forcing the network to seek other relevant parts when the most discriminative part is hidden. Our approach only needs to modify the input image and can work with any network designed for object localization. During testing, we do not need to hide any patches. Our Hide-and-Seek approach obtains superior performance compared to previous methods for weakly-supervised object localization on the ILSVRC dataset. We also demonstrate that our framework can be easily extended to weakly-supervised action localization.

Learning to hash has been widely applied to approximate nearest neighbor search for large-scale multimedia retrieval, due to its computation efficiency and retrieval quality. Deep learning to hash, which improves retrieval quality by end-to-end representation learning and hash encoding, has received increasing attention recently. Subject to the ill-posed gradient difficulty in the optimization with sign activations, existing deep learning to hash methods need to first learn continuous representations and then generate binary hash codes in a separated binarization step, which suffer from substantial loss of retrieval quality. This work presents HashNet, a novel deep architecture for deep learning to hash by continuation method with convergence guarantees, which learns exactly binary hash codes from imbalanced similarity data. The key idea is to attack the ill-posed gradient problem in optimizing deep networks with non-smooth binary activations by continuation method, in which we begin from learning an easier network with smoothed activation function and let it evolve during the training, until it eventually goes back to being the original, difficult to optimize, deep network with the sign activation function. Comprehensive empirical evidence shows that HashNet can generate exactly binary hash codes and yield state-of-the-art multimedia retrieval performance on standard benchmarks.

Convolutional Neural Networks have been a subject of great importance over the past decade and great strides have been made in their utility for producing state of the art performance in many computer vision problems. However, the behavior of deep networks is yet to be fully understood and is still an active area of research. In this work, we present an intriguing behavior: pre-trained CNNs can be made to improve their predictions by structurally perturbing the input. We observe that these perturbations - referred as Guided Perturbations - enable a trained network to improve its prediction performance without any learning or change in network weights. We perform various ablative experiments to understand how these perturbations affect the local context and feature representations. Furthermore, we demonstrate that this idea can improve performance of several existing approaches on semantic segmentation and scene labeling tasks on the PASCAL VOC dataset and supervised classification tasks on MNIST and CIFAR10 datasets.

Person re-identification is best known as the problem of associating a single person that is observed from one or more disjoint cameras. The existing literature has mainly addressed such an issue, neglecting the fact that people usually move in groups, like in crowded scenarios. We believe that the additional information carried by neighboring individuals provides a relevant visual context that can be exploited to obtain a more robust match of single persons within the group. Despite this, re-identifying groups of people compound the common single person re-identification problems by introducing changes in the relative position of persons within the group and severe self-occlusions. In this paper, we propose a solution for group re-identification that grounds on transferring knowledge from single person re-identification to group re-identification by exploiting sparse dictionary learning. First, a dictionary of sparse atoms is learned using patches extracted from single person images. Then, the learned dictionary is exploited to obtain a sparsity-driven residual group representation, which is finally matched to perform the re-identification. Extensive experiments on the i-LIDS groups and two newly collected datasets show that the proposed solution outperforms state-of-the-art approaches.

We propose a technique for producing 'visual explanations' for decisions from a large class of Convolutional Neural Network (CNN)-based models, making them more transparent. Our approach - Gradient-weighted Class Activation Mapping (Grad-CAM), uses the gradients of any target concept (say logits for 'dog' or even a caption), flowing into the final convolutional layer to produce a coarse localization map highlighting the important regions in the image for predicting the concept. Unlike previous approaches, Grad-CAM is applicable to a wide variety of CNN model-families: (1) CNNs with fully-connected layers (e.g. VGG), (2) CNNs used for structured outputs (e.g. captioning), (3) CNNs used in tasks with multi-modal inputs (e.g. VQA) or reinforcement learning, and needs no architectural changes or re-training. We combine Grad-CAM with existing fine-grained visualizations to create a high-resolution class-discriminative visualization and apply it to image classification, image captioning, and visual question answering (VQA) models, including ResNet-based architectures. In the context of image classification models, our visualizations (a) lend insights into failure modes of these models (showing that seemingly unreasonable predictions have reasonable explanations), (b) outperform previous methods on the ILSVRC-15 weakly-supervised localization task, (c) are more faithful to the underlying model, and (d) help achieve model generalization by identifying dataset bias. For image captioning and VQA, our visualizations show that even non-attention based models can localize inputs. Finally, we design and conduct human studies to measure if Grad-CAM explanations help users establish appropriate trust in predictions from deep networks and show that Grad-CAM helps untrained users successfully discern a 'stronger' deep network from a 'weaker' one even when both make identical predictions. Our code is available at https://github.com/ramprs/grad-cam/ along with a demo on CloudCV [2] 1 and video at youtu.be/COjUB9Izk6E.

Most present-day imaging devices are equipped with CMOS sensors. Motion blur is a common artifact in hand-held cameras. Because CMOS sensors mostly employ a rolling shutter (RS), the motion deblurring problem takes on a new dimension. Although few works have recently addressed this problem, they suffer from many constraints including heavy computational cost, need for precise sensor information, and inability to deal with wide-angle systems (which most cell-phone and drone cameras are) and irregular camera trajectory. In this work, we propose a model for RS blind motion deblurring that mitigates these issues significantly. Comprehensive comparisons with state-of-the-art methods reveal that our approach not only exhibits significant computational gains and unconstrained functionality but also leads to improved deblurring performance.

The deep convolutional neural network (CNN) is the state-of-the-art solution for large-scale visual recognition. Following some basic principles such as increasing network depth and constructing highway connections, researchers have manually designed a lot of fixed network architectures and verified their effectiveness. In this paper, we discuss the possibility of learning deep network structures automatically. Note that the number of possible network structures increases exponentially with the number of layers in the network, which motivates us to adopt the genetic algorithm to efficiently explore this large search space. The core idea is to propose an encoding method to represent each network structure in a fixed-length binary string. The genetic algorithm is initialized by generating a set of randomized individuals. In each generation, we define standard genetic operations, e.g., selection, mutation and crossover, to generate competitive individuals and eliminate weak ones. The competitiveness of each individual is defined as its recognition accuracy, which is obtained via a standalone training process on a reference dataset. We run the genetic process on CIFAR10, a small-scale dataset, demonstrating its ability to find high-quality structures which are little studied before. The learned powerful structures are also transferrable to the ILSVRC2012 dataset for large-scale visual recognition.

Humans infer rich knowledge of objects from both auditory and visual cues. Building a machine of such competency, however, is very challenging, due to the great difficulty in capturing large-scale, clean data of objects with both their appearance and the sound they make. In this paper, we present a novel, open-source pipeline that generates audio-visual data, purely from 3D object shapes and their physical properties. Through comparison with audio recordings and human behavioral studies, we validate the accuracy of the sounds it generates. Using this generative model, we are able to construct a synthetic audio-visual dataset, namely Sound-20K, for object perception tasks. We demonstrate that auditory and visual information play complementary roles in object perception, and further, that the representation learned on synthetic audio-visual data can transfer to real-world scenarios.

Recent advancements in generative adversarial networks (GANs), using deep convolutional models, have supported the development of image generation techniques able to reach satisfactory levels of realism. Further improvements have been proposed to condition GANs to generate images matching a specific object category or a short text description. In this work, we build on the latter class of approaches and investigate the possibility of driving and conditioning the image generation process by means of brain signals recorded, through an electroencephalograph (EEG), while users look at images from a set of 40 ImageNet object categories with the objective of generating the seen images. To accomplish this task, we first demonstrate that brain activity EEG signals encode visually-related information that allows us to accurately discriminate between visual object categories and, accordingly, we extract a more compact class-dependent representation of EEG data using recurrent neural networks. Afterwards, we use the learned EEG manifold to condition image generation employing GANs, which, during inference, will read EEG signals and convert them into images. We tested our generative approach using EEG signals recorded from six subjects while looking at images of the aforementioned 40 visual classes. The results show that for classes represented by well-defined visual patterns (e.g., pandas, airplane, etc.), the generated images are realistic and highly resemble those evoking the EEG signals used for conditioning GANs, resulting in an actual reading-the-mind process.

We present a novel method called Contextual Pyramid CNN (CP-CNN) for generating high-quality crowd density and count estimation by explicitly incorporating global and local contextual information of crowd images. The proposed CP-CNN consists of four modules: Global Context Estimator (GCE), Local Context Estimator (LCE), Density Map Estimator (DME) and a Fusion-CNN (F-CNN). GCE is a VGG-16 based CNN that encodes global context and it is trained to classify input images into different density classes, whereas LCE is another CNN that encodes local context information and it is trained to perform patch-wise classification of input images into different density classes. DME is a multi-column architecture-based CNN that aims to generate high-dimensional feature maps from the input image which are fused with the contextual information estimated by GCE and LCE using F-CNN. To generate high resolution and high-quality density maps, F-CNN uses a set of convolutional and fractionally-strided convolutional layers and it is trained along with the DME in an end-to-end fashion using a combination of adversarial loss and pixel-level Euclidean loss. Extensive experiments on highly challenging datasets show that the proposed method achieves significant improvements over the state-of-the-art methods.

Most recent CNN architectures use average pooling as a final feature encoding step. In the field of fine-grained recognition, however, recent global representations like bilinear pooling offer improved performance. In this paper, we generalize average and bilinear pooling to "alpha-pooling", allowing for learning the pooling strategy during training. In addition, we present a novel way to visualize decisions made by these approaches. We identify parts of training images having the highest influence on the prediction of a given test image. This allows for justifying decisions to users and also for analyzing the influence of semantic parts. For example, we can show that the higher capacity VGG16 model focuses much more on the bird's head than, e.g., the lower-capacity VGG-M model when recognizing fine-grained bird categories. Both contributions allow us to analyze the difference when moving between average and bilinear pooling. In addition, experiments show that our generalized approach can outperform both across a variety of standard datasets.

We tackle the problem of learning robotic sensorimotor control policies that can generalize to visually diverse and unseen environments. Achieving broad generalization typically requires large datasets, which are difficult to obtain for task-specific interactive processes such as reinforcement learning or learning from demonstration. However, much of the visual diversity in the world can be captured through passively collected datasets of images or videos. In our method, which we refer to as GPLAC (Generalized Policy Learning with Attentional Classifier), we use both interaction data and weakly labeled image data to augment the generalization capacity of sensorimotor policies. Our method combines multitask learning on action selection and an auxiliary binary classification objective, together with a convolutional neural network architecture that uses an attentional mechanism to avoid distractors. We show that pairing interaction data from just a single environment with a diverse dataset of weakly labeled data results in greatly improved generalization to unseen environments, and show that this generalization depends on both the auxiliary objective and the attentional architecture that we propose. We demonstrate our results in both simulation and on a real robotic manipulator, and demonstrate substantial improvement over standard convolutional architectures and domain adaptation methods.

Research into computational jigsaw puzzle solving, an emerging theoretical problem with numerous applications, has focused in recent years on puzzles that constitute square pieces only. In this paper we wish to extend the scientific scope of appearance-based puzzle solving and consider "brick wall" jigsaw puzzles - rectangular pieces who may have different sizes, and could be placed next to each other at arbitrary offset along their abutting edge -- a more explicit configuration with propertie of real world puzzles. We present the new challenges that arise in brick wall puzzles and address them in two stages. First we concentrate on the reconstruction of the puzzle (with or without missing pieces) assuming an oracle for offset assignments. We show that despite the increased complexity of the problem, under these conditions performance can be made comparable to the state-of-the-art in solving the simpler square piece puzzles, and thereby argue that solving brick wall puzzles may be reduced to finding the correct offset between two neighboring pieces. We then move on to focus on implementing the oracle computationally using a mixture of dissimilarity metrics and correlation matching. We show results on various brick wall puzzles and discuss how our work may start a new research path for the puzzle solving community.

Spectral analysis of natural scenes can provide much more detailed information about the scene than an ordinary RGB camera. The richer information provided by hyperspectral images has been beneficial to numerous applications, such as understanding natural environmental changes and classifying plants and soils in agriculture based on their spectral properties. In this paper, we present an efficient manifold learning based method for accurately reconstructing a hyperspectral image from a single RGB image captured by a commercial camera with known spectral response. By applying a nonlinear dimensionality reduction technique to a large set of natural spectra, we show that the spectra of natural scenes lie on an intrinsically low dimensional manifold. This allows us to map an RGB vector to its corresponding hyperspectral vector accurately via our proposed novel manifold-based reconstruction pipeline. Experiments using both synthesized RGB images using hyperspectral datasets and real world data demonstrate our method outperforms the state-of-the-art.