CVPR 2015 Papers

In this paper we present a symmetric KL divergence based agglomerative clustering framework to segment multiple levels of depigmentation in Vitiligo images. The proposed framework starts with a simple merge cost based on symmetric KL divergence. We extend the recent body of work related to Bregman divergence based agglomerative clustering and prove that the symmetric KL divergence is an upper-bound for uni-modal Gaussian distributions. This leads to a very simple yet elegant method for bottomup agglomerative clustering. We introduce albedo and reflectance fields as features for the distance computations. We compare against other established methods to bring out possible pros and cons of the proposed method.

We tackle a fundamental problem to detect and estimate just noticeable blur (JNB) caused by defocus that spans a small number of pixels in images. This type of blur is common during photo taking. Although it is not strong, the slight edge blurriness contains informative clues related to depth. We found existing blur descriptors based on local information cannot distinguish this type of small blur reliably from unblurred structures. We propose a simple yet effective blur feature via sparse representation and image decomposition. It directly establishes correspondence between sparse edge representation and blur strength estimation. Extensive experiments manifest the generality and robustness of this feature.

In this paper, we focus on heterogeneous feature learning for RGB-D activity recognition. Considering that features from different channels could share some similar hidden structures, we propose a joint learning model to simultaneously explore the shared and feature-specific components as an instance of heterogenous multi-task learning. The proposed model in an unified framework is capable of: 1) jointly mining a set of subspaces with the same dimensionality to enable the multi-task classifier learning, and 2) meanwhile, quantifying the shared and feature-specific components of features in the subspaces. To efficiently train the joint model, a three-step iterative optimization algorithm is proposed, followed by two inference models. Extensive results on three activity datasets have demonstrated the efficacy of the proposed method. In addition, a novel RGB-D activity dataset focusing on human-object interaction is collected for evaluating the proposed method, which will be made available to the community for RGB-D activity benchmarking and analysis.

We present a novel vanishing point (VP) detection and tracking algorithm for calibrated monocular image sequences. Previous VP detection and tracking methods usually assume known camera poses for all frames or detect and track separately. We advance the state-of-the-art by combining VP extraction on a Gaussian sphere with recent advances in multi-target tracking on probabilistic occupancy fields. The solution is obtained by solving a Linear Program (LP). This enables the joint detection and tracking of multiple VPs over sequences. Unlike existing works we do not need known camera poses, and at the same time avoid detecting and tracking in separate steps. We also propose an extension to enforce VP orthogonality. We augment an existing video dataset consisting of 48 monocular videos with multiple annotated VPs in 14448 frames for evaluation. Although the method is designed for unknown camera poses, it is also helpful in scenarios with known poses, since a multi-frame approach in VP detection helps to regularize in frames with weak VP line support.

Tracking-by-detection has proven to be the most successful strategy to address the task of tracking multiple targets in unconstrained scenarios. Traditionally, a set of sparse detections, generated in a preprocessing step, serves as input to a high-level tracker whose goal is to correctly associate these "dots" over time. An obvious shortcoming of this approach is that most information available in image sequences is simply ignored by thresholding weak detection responses and applying non-maximum suppression. We propose a multi-target tracker that exploits low level image information and associates every (super)-pixel to a specific target or classifies it as background. As a result, we obtain an video segmentation in addition to the classical bounding-box representation in unconstrained, real-world sequences. Our method shows encouraging results on many standard benchmark sequences and significantly outperforms state-of-the-art tracking-by-detection approaches in crowded scenes with long-term partial occlusions.

We present a system for fast and highly accurate 3D localization of objects like cars in autonomous driving applications, using a single camera. Our localization framework jointly uses information from complementary modalities such as structure from motion (SFM) and object detection to achieve high localization accuracy in both near and far fields. This is in contrast to prior works that rely purely on detector outputs, or motion segmentation based on sparse feature tracks. Rather than completely commit to tracklets generated by a 2D tracker, we make novel use of raw detection scores to allow our 3D bounding boxes to adapt to better quality 3D cues. To extract SFM cues, we demonstrate the advantages of dense tracking over sparse mechanisms in autonomous driving scenarios. In contrast to complex scene understanding, our formulation for 3D localization is efficient and can be regarded as an extension of sparse bundle adjustment to incorporate object detection cues. Experiments on the KITTI dataset show the efficacy of our cues, as well as the accuracy and robustness of our 3D object localization relative to ground truth and prior works.

We propose an approach that utilizes large collections of photo streams and blog posts, two of the most prevalent sources of data on the Web, for joint story-based summarization and exploration. Blogs consist of sequences of images and associated text; they portray events and experiences with concise sentences and representative images. We leverage blogs to help achieve story-based semantic summarization of collections of photo streams. In the opposite direction, blog posts can be enhanced with sets of photo streams by showing interpolations between consecutive images in the blogs. We formulate the problem of joint alignment from blogs to photo streams and photo stream summarization in a unified latent ranking SVM framework. We alternate between solving the two coupled latent SVM problems, by first fixing the summarization and solving for the alignment from blog images to photo streams and vice versa. On a newly collected large-scale Disneyland dataset of 10K blogs (120K associated images) and 6K photo streams (540K images), we demonstrate that blog posts and photo streams are mutually beneficial for summarization, exploration, semantic knowledge transfer, and photo interpolation.

The face is one of the most powerful channel of non-verbal communication. The most commonly used taxonomy to describe facial behaviour is the Facial Action Coding System (FACS). FACS segments the visible effects of facial muscle activation into 30+ action units (AUs). AUs, which may occur alone and in thousands of combinations, can describe nearly all-possible facial expressions. Most existing methods for automatic AU detection treat the problem using one-vs-all classifiers and fail to exploit dependencies among AU and facial features. We introduce joint-patch and multi-label learning (JPML) to address these issues. JPML leverages group sparsity by selecting a sparse subset of facial patches while learning a multi-label classifier. In four of five comparisons on three diverse datasets, CK+, GFT, and BP4D, JPML produced the highest average F1 scores in comparison with state-of-the art.

In the semantic multinomial framework patches and images are modeled as points in a semantic probability simplex. Patch theme models are learned resorting to weak supervision via image labels, which leads the problem of scene categories co-occurring in this semantic space. Fortunately, each category has its own co-occurrence patterns that are consistent across the images in that category. Thus, discovering and modeling these patterns is critical to improve the recognition performance in this representation. In this paper, we observe that not only global co-occurrences at the image-level are important, but also different regions have different category co-occurrence patterns. We exploit local contextual relations to address the problem of discovering consistent co-occurrence patterns and removing noisy ones. Our hypothesis is that a less noisy semantic representation, would greatly help the classifier to model consistent co-occurrences and discriminate better between scene categories. An important advantage of modeling features in a semantic space, is that this space is feature independent. Thus, we can combine multiple features and spatial neighbors in the same common space, and formulate the problem as minimizing a context-dependent energy. Experimental results show that exploiting different types of contextual relations consistently improves the recognition accuracy. In particular, larger datasets benefit more from the proposed method, leading to very competitive performance.

With the advent of drones, aerial video analysis becomes increasingly important; yet, it has received scant attention in the literature. This paper addresses a new problem of parsing low-resolution aerial videos of large spatial areas, in terms of 1) grouping, 2) recognizing events and 3) assigning roles to people engaged in events. We propose a novel framework aimed at conducting joint inference of the above tasks, as reasoning about each in isolation typically fails in our setting. Given noisy tracklets of people and detections of large objects and scene surfaces (e.g., building, grass), we use a spatiotemporal AND-OR graph to drive our joint inference, using Markov Chain Monte Carlo and dynamic programming. We also introduce a new formalism of spatiotemporal templates characterizing latent sub-events. For evaluation, we have collected and released a new aerial videos dataset using a hex-rotor flying over picnic areas rich with group events. Our results demonstrate that we successfully address above inference tasks under challenging conditions.

We present a method for calibrating the Ensemble of Exemplar SVMs model. Unlike the standard approach, which calibrates each SVM independently, our method optimizes their joint performance as an ensemble. We formulate joint calibration as a constrained optimization problem and devise an efficient optimization algorithm to find its global optimum. The algorithm dynamically discards parts of the solution space that cannot contain the optimum early on, making the optimization computationally feasible. We experiment with EE-SVM trained on state-of-the-art CNN descriptors. Results on the ILSVRC 2014 and PASCAL VOC 2007 datasets show that (i) our joint calibration procedure outperforms independent calibration on the task of classifying windows as belonging to an object class or not; and (ii) this improved window classifier leads to better performance on the object detection task.

Action recognition and pose estimation from video are closely related tasks for understanding human motion, most methods, however, learn separate models and combine them sequentially. In this paper, we propose a framework to integrate training and testing of the two tasks. A spatial- temporal And-Or graph model is introduced to represent action at three scales. Specifically the action is decomposed into poses which are further divided to mid-level ST-parts and then parts. The hierarchical structure of our model captures the geometric and appearance variations of pose at each frame and lateral connections between ST-parts at adjacent frames capture the action-specific motion information. The model parameters for three scales are learned discriminatively, and action labels and poses are efficiently inferred by dynamic programming. Experiments demonstrate that our approach achieves state-of-art accuracy in action recognition while also improving pose estimation.

We present a novel Joint Online Tracking and Segmentation (JOTS) algorithm which integrates the multi-part tracking and segmentation into a unified energy optimization framework to handle the video segmentation task. The multi-part segmentation is posed as a pixel-level label assignment task with regularization according to the estimated part models, and tracking is formulated as estimating the part models based on the pixel labels, which in turn is used to refine the model. The multi-part tracking and segmentation are carried out iteratively to minimize the proposed objective function by a RANSAC-style approach. Extensive experiments on the SegTrack and SegTrack v2 databases demonstrate that the proposed algorithm performs favorably against the state-of-the-art methods.

While widely acknowledged as highly effective in computer vision, multi-label MRFs with non-convex priors are difficult to optimize. To tackle this, we introduce an algorithm that iteratively approximates the original energy with an appropriately weighted surrogate energy that is easier to minimize. Our algorithm guarantees that the original energy decreases at each iteration. In particular, we consider the scenario where the global minimizer of the weighted surrogate energy can be obtained by a multi-label graph cut algorithm, and show that our algorithm then lets us handle of large variety of non-convex priors. We demonstrate the benefits of our method over state-of-the-art MRF energy minimization techniques on stereo and inpainting problems.

Successful visual object recognition methods typically rely on training datasets containing lots of richly annotated images. Annotating object bounding boxes is both expensive and subjective. We describe a weakly supervised convolutional neural network (CNN) for object classification that relies only on image-level labels, yet can learn from cluttered scenes containing multiple objects. We quantify its object classification and object location prediction performance on the Pascal VOC 2012 (20 object classes) and the much larger Microsoft COCO (80 object classes) datasets. We find that the network (i) outputs accurate image-level labels, (ii) predicts approximate locations (but not extents) of objects, and (iii) performs similar or better compared to its fully-supervised counterparts using object bounding box annotation for training.

This work presents a novel approach for detecting inliers in a given set of correspondences (matches). It does so without explicitly identifying any consensus set, based on a method for inlier rate estimation (IRE). Given such an estimator for the inlier rate, we also present an algorithm that detects a globally optimal transformation. We provide a theoretical analysis of the IRE method using a stochastic generative model on the continuous spaces of matches and transformations. This model allows rigorous investigation of the limits of our IRE method for the case of 2D-translation, further giving bounds and insights for the more general case. Our theoretical analysis is validated empirically and is shown to hold in practice for the more general case of 2D-affinities. In addition, we show that the combined framework works on challenging cases of 2D-homography estimation, with very few and possibly noisy inliers, where RANSAC generally fails.

Camera motion introduces motion blur, degrading the quality of video. A video deblurring method is proposed based on two observations: (i) camera motion within capture of each individual frame leads to motion blur; (ii) camera motion between frames yields inter-frame mis-alignment that can be exploited for blur removal. The proposed method effectively leverages the information distributed across multiple video frames due to camera motion, jointly estimating the motion between consecutive frames and blur within each frame. This joint analysis is crucial for achieving effective restoration by leveraging temporal information. Extensive experiments are carried out on synthetic data as well as real-world blurry videos. Comparisons with several state-of-the-art methods verify the effectiveness of the proposed method.

Despite tremendous progress in computer vision, effective learning on very large-scale (>100K patients) medical image databases has been vastly hindered. We present an interleaved text/image deep learning system to extract and mine the semantic interactions of radiology images and reports from a national research hospital's picture archiving and communication system. Instead of using full 3D medical volumes, we focus on a collection of representative ~216K 2D key images/slices (selected by clinicians for diagnostic reference) with text-driven scalar and vector labels. Our system interleaves between unsupervised learning (e.g., latent Dirichlet allocation, recurrent neural net language models) on document- and sentence-level texts to generate semantic labels and supervised learning via deep convolutional neural networks (CNNs) to map from images to label spaces. Disease-related key words can be predicted for radiology images in a retrieval manner. We have demonstrated promising quantitative and qualitative results. The large-scale datasets of extracted key images and their categorization, embedded vector labels and sentence descriptions can be harnessed to alleviate the deep learning "data-hungry" obstacle in the medical domain.

Modeling human-object interactions and manipulating motions lies in the heart of fine-grained action recognition. Previous methods heavily rely on explicit detection of the object being interacted, which requires intensive human labour on object annotation. To bypass this constraint and achieve better classification performance, in this work, we propose a novel fine-grained action recognition pipeline by interaction part proposal and discriminative mid-level part mining. Firstly, we generate a large number of candidate object regions using off-the-shelf object proposal tool, e.g., BING. Secondly, these object regions are matched and tracked across frames to form a large spatio-temporal graph based on the appearance matching and the dense motion trajectories through them. We then propose an efficient approximate graph segmentation algorithm to partition and filter the graph into consistent local dense sub-graphs. These sub-graphs, which are spatio-temporal sub-volumes, represent our candidate interaction parts. Finally, we mine discriminative mid-level part detectors from the features computed over the candidate interaction parts. Bag-of-detection scores based on a novel Max-N pooling scheme are computed as the action representation for a video sample. We conduct extensive experiments on human-object interaction datasets including MPII Cooking and MSR Daily Activity 3D. The experimental results demonstrate that the proposed framework achieves consistent improvements over the state-of-the-art action recognition accuracies on the benchmarks, without using any object annotation.