Simon Hadfield

The Visual Object Tracking VOT2014 Challenge Results

Visual tracking has attracted a significant attention in the last few decades. The recent surge in the number of publications on tracking-related problems have made it almost impossible to follow the developments in the field. One of the reasons is that there is a lack of commonly accepted annotated data-sets and standardized evaluation protocols that would allow objective comparison of different tracking methods. To address this issue, the Visual Object Tracking (VOT) workshop was organized in conjunction with ICCV2013. Researchers from academia as well as industry were invited to participate in the first VOT2013 challenge which aimed at single-object visual trackers that do not apply pre-learned models of object appearance (model-free). Presented here is the VOT2013 benchmark dataset for evaluation of single-object visual trackers as well as the results obtained by the trackers competing in the challenge. In contrast to related attempts in tracker benchmarking, the dataset is labeled per-frame by visual attributes that indicate occlusion, illumination change, motion change, size change and camera motion, offering a more systematic comparison of the trackers. Furthermore, we have designed an automated system for performing and evaluating the experiments. We present the evaluation protocol of the VOT2013 challenge and the results of a comparison of 27 trackers on the benchmark dataset. The dataset, the evaluation tools and the tracker rankings are publicly available from the challenge website (http://votchallenge.net).

Hollywood 3D: Recognizing Actions in 3D Natural Scenes

Action recognition in unconstrained situations is a difficult task, suffering from massive intra-class variations. It is made even more challenging when complex 3D actions are projected down to the image plane, losing a great deal of information. The recent emergence of 3D data, both in broadcast content, and commercial depth sensors, provides the possibility to overcome this issue. This paper presents a new dataset, for benchmarking action recognition algorithms in natural environments, while making use of 3D information. The dataset contains around 650 video clips, across 14 classes. In addition, two state of the art action recognition algorithms are extended to make use of the 3D data, and five new interest point detection strategies are also proposed, that extend to the 3D data. Our evaluation compares all 4 feature descriptors, using 7 different types of interest point, over a variety of threshold levels, for the Hollywood3D dataset. We make the dataset including stereo video, estimated depth maps and all code required to reproduce the benchmark results, available to the wider community.

Kinecting the dots: Particle based scene flow from depth sensors

The motion field of a scene can be used for object segmentation and to provide features for classification tasks like action recognition. Scene flow is the full 3D motion field of the scene, and is more difficult to estimate than its 2D counterpart, optical flow. Current approaches use a smoothness cost for regularisation, which tends to over-smooth at object boundaries. This paper presents a novel formulation for scene flow estimation, a collection of moving points in 3D space, modelled using a particle filter that supports multiple hypotheses and does not oversmooth the motion field. In addition, this paper is the first to address scene flow estimation, while making use of modern depth sensors and monocular appearance images, rather than traditional multi-viewpoint rigs. The algorithm is applied to an existing scene flow dataset, where it achieves comparable results to approaches utilising multiple views, while taking a fraction of the time.

Scene Particles: Unregularized Particle-Based Scene Flow Estimation

In this paper, an algorithm is presented for estimating scene flow, which is a richer, 3D analog of optical flow. The approach operates orders of magnitude faster than alternative techniques and is well suited to further performance gains through parallelized implementation. The algorithm employs multiple hypotheses to deal with motion ambiguities, rather than the traditional smoothness constraints, removing oversmoothing errors and providing significant performance improvements on benchmark data, over the previous state of the art. The approach is flexible and capable of operating with any combination of appearance and/or depth sensors, in any setup, simultaneously estimating the structure and motion if necessary. Additionally, the algorithm propagates information over time to resolve ambiguities, rather than performing an isolated estimation at each frame, as in contemporary approaches. Approaches to smoothing the motion field without sacrificing the benefits of multiple hypotheses are explored, and a probabilistic approach to occlusion estimation is demonstrated, leading to 10 and 15 percent improved performance, respectively. Finally, a data-driven tracking approach is described, and used to estimate the 3D trajectories of hands during sign language, without the need to model complex appearance variations at each viewpoint.

Long-Term Tracking through Failure Cases

Long term tracking of an object, given only a single instance in an initial frame, remains an open problem. We propose a visual tracking algorithm, robust to many of the difficulties which often occur in real-world scenes. Correspondences of edge-based features are used, to overcome the reliance on the texture of the tracked object and improve invariance to lighting. Furthermore we address long-term stability, enabling the tracker to recover from drift and to provide redetection following object disappearance or occlusion. The two-module principle is similar to the successful state-of-the-art long-term TLD tracker, however our approach extends to cases of low-textured objects. Besides reporting our results on the VOT Challenge dataset, we perform two additional experiments. Firstly, results on short-term sequences show the performance of tracking challenging objects which represent failure cases for competing state-of-the-art approaches. Secondly, long sequences are tracked, including one of almost 30000 frames which to our knowledge is the longest tracking sequence reported to date. This tests the re-detection and drift resistance properties of the tracker. All the results are comparable to the state-of-the-art on sequences with textured objects and superior on non-textured objects. The new annotated sequences are made publicly available.

Natural Action Recognition Using Invariant 3D Motion Encoding

We investigate the recognition of actions “in the wild” using 3D motion information. The lack of control over (and knowledge of) the camera configuration, exacerbates this already challenging task, by introducing systematic projective inconsistencies between 3D motion fields, hugely increasing intra-class variance. By introducing a robust, sequence based, stereo calibration technique, we reduce these inconsistencies from fully projective to a simple similarity transform. We then introduce motion encoding techniques which provide the necessary scale invariance, along with additional invariances to changes in camera viewpoint.

Using Convolutional 3D Neural Networks for User-independent continuous gesture recognition

In this paper, we propose using 3D Convolutional Neural Networks for large scale user-independent continuous gesture recognition. We have trained an end-to-end deep network for continuous gesture recognition (jointly learning both the feature representation and the classifier). The network performs three-dimensional (i.e. space-time) convolutions to extract features related to both the appearance and motion from volumes of color frames. Space-time invariance of the extracted features is encoded via pooling layers. The earlier stages of the network are partially initialized using the work of Tran et al. before being adapted to the task of gesture recognition. An earlier version of the proposed method, which was trained for 11,250 iterations, was submitted to ChaLearn 2016 Continuous Gesture Recognition Challenge and ranked 2nd with the Mean Jaccard Index Score of 0.269235. When the proposed method was further trained for 28,750 iterations, it achieved state-of-the-art performance on the same dataset, yielding a 0.314779 Mean Jaccard Index Score.

2D or Not 2D: Bridging the Gap Between Tracking and Structure from Motion

In this paper, we address the problem of tracking an unknown object in 3D space. Online 2D tracking often fails for strong out-of-plane rotation which results in considerable changes in appearance beyond those that can be represented by online update strategies. However, by modelling and learning the 3D structure of the object explicitly, such effects are mitigated. To address this, a novel approach is presented, combining techniques from the fields of visual tracking, structure from motion (SfM) and simultaneous localisation and mapping (SLAM). This algorithm is referred to as TMAGIC (Tracking, Modelling And Gaussian-process Inference Combined). At every frame, point and line features are tracked in the image plane and are used, together with their 3D correspondences, to estimate the camera pose. These features are also used to model the 3D shape of the object as a Gaussian process. Tracking determines the trajectories of the object in both the image plane and 3D space, but the approach also provides the 3D object shape. The approach is validated on several video-sequences used in the tracking literature, comparing favourably to state-of-the-art trackers for simple scenes (error reduced by 22 %) with clear advantages in the case of strong out-of-plane rotation, where 2D approaches fail (error reduction of 58 %).

Texture-Independent Long-Term Tracking Using Virtual Corners

Long-term tracking of an object, given only a single instance in an initial frame, remains an open problem. We propose a visual tracking algorithm, robust to many of the difficulties that often occur in real-world scenes. Correspondences of edge-based features are used, to overcome the reliance on the texture of the tracked object and improve invariance to lighting. Furthermore, we address long-term stability, enabling the tracker to recover from drift and to provide redetection following object disappearance or occlusion. The two-module principle is similar to the successful state-of-the-art long-term TLD tracker; however, our approach offers better performance in benchmarks and extends to cases of low-textured objects. This becomes obvious in cases of plain objects with no texture at all, where the edge-based approach proves the most beneficial. We perform several different experiments to validate the proposed method. First, results on short-term sequences show the performance of tracking challenging (low textured and/or transparent) objects that represent failure cases for competing the state-of-the-art approaches. Second, long sequences are tracked, including one of almost 30 000 frames, which, to the best of our knowledge, is the longest tracking sequence reported to date. This tests the redetection and drift resistance properties of the tracker. Finally, we report the results of the proposed tracker on the VOT Challenge 2013 and 2014 data sets as well as on the VTB1.0 benchmark, and we show relative performance of the tracker compared with its competitors. All the results are comparable with the state of the art on sequences with textured objects and superior on non-textured objects. The new annotated sequences are made publicly available.

Generalised pose estimation using depth

Estimating the pose of an object, be it articulated, deformable or rigid, is an important task, with applications ranging from Human-Computer Interaction to environmental understanding. The idea of a general pose estimation framework, capable of being rapidly retrained to suit a variety of tasks, is appealing. In this paper a solution is proposed requiring only a set of labelled training images in order to be applied to many pose estimation tasks. This is achieved by treating pose estimation as a classification problem, with particle filtering used to provide non-discretised estimates. Depth information extracted from a calibrated stereo sequence, is used for background suppression and object scale estimation. The appearance and shape channels are then transformed to Local Binary Pattern histograms, and pose classification is performed via a randomised decision forest. To demonstrate flexibility, the approach is applied to two different situations, articulated hand pose and rigid head orientation, achieving 97% and 84% accurate estimation rates, respectively.