Header logo is


2018


Learning 3D Shape Completion under Weak Supervision
Learning 3D Shape Completion under Weak Supervision

Stutz, D., Geiger, A.

Arxiv, May 2018 (article)

Abstract
We address the problem of 3D shape completion from sparse and noisy point clouds, a fundamental problem in computer vision and robotics. Recent approaches are either data-driven or learning-based: Data-driven approaches rely on a shape model whose parameters are optimized to fit the observations; Learning-based approaches, in contrast, avoid the expensive optimization step by learning to directly predict complete shapes from incomplete observations in a fully-supervised setting. However, full supervision is often not available in practice. In this work, we propose a weakly-supervised learning-based approach to 3D shape completion which neither requires slow optimization nor direct supervision. While we also learn a shape prior on synthetic data, we amortize, i.e., learn, maximum likelihood fitting using deep neural networks resulting in efficient shape completion without sacrificing accuracy. On synthetic benchmarks based on ShapeNet and ModelNet as well as on real robotics data from KITTI and Kinect, we demonstrate that the proposed amortized maximum likelihood approach is able to compete with fully supervised baselines and outperforms data-driven approaches, while requiring less supervision and being significantly faster.

avg

PDF Project Page Project Page [BibTex]


no image
Schema-related cognitive load influences performance, speech, and physiology in a dual-task setting: A continuous multi-measure approach

Wirzberger, M., Herms, R., Esmaeili Bijarsari, S., Eibl, M., Rey, G. D.

Cognitive Research: Principles and Implications, 3:46, Springer Nature, 2018 (article)

Abstract
Schema acquisition processes comprise an essential source of cognitive demands in learning situations. To shed light on related mechanisms and influencing factors, this study applied a continuous multi-measure approach for cognitive load assessment. In a dual-task setting, a sample of 123 student participants learned visually presented symbol combinations with one of two levels of complexity while memorizing auditorily presented number sequences. Learners’ cognitive load during the learning task was addressed by secondary task performance, prosodic speech parameters (pauses, articulation rate), and physiological markers (heart rate, skin conductance response). While results revealed increasing primary and secondary task performance over the trials, decreases in speech and physiological parameters indicated a reduction in the overall level of cognitive load with task progression. In addition, the robustness of the acquired schemata was confirmed by a transfer task that required participants to apply the obtained symbol combinations. Taken together, the observed pattern of evidence supports the idea of a logarithmically decreasing progression of cognitive load with increasing schema acquisition, and further hints on robust and stable transfer performance, even under enhanced transfer demands. Finally, theoretical and practical consequences consider evidence on desirable difficulties in learning as well as the potential of multimodal cognitive load detection in learning applications.

re

DOI [BibTex]

DOI [BibTex]


Augmented Reality Meets Computer Vision: Efficient Data Generation for Urban Driving Scenes
Augmented Reality Meets Computer Vision: Efficient Data Generation for Urban Driving Scenes

Alhaija, H., Mustikovela, S., Mescheder, L., Geiger, A., Rother, C.

International Journal of Computer Vision (IJCV), 2018, 2018 (article)

Abstract
The success of deep learning in computer vision is based on the availability of large annotated datasets. To lower the need for hand labeled images, virtually rendered 3D worlds have recently gained popularity. Unfortunately, creating realistic 3D content is challenging on its own and requires significant human effort. In this work, we propose an alternative paradigm which combines real and synthetic data for learning semantic instance segmentation and object detection models. Exploiting the fact that not all aspects of the scene are equally important for this task, we propose to augment real-world imagery with virtual objects of the target category. Capturing real-world images at large scale is easy and cheap, and directly provides real background appearances without the need for creating complex 3D models of the environment. We present an efficient procedure to augment these images with virtual objects. In contrast to modeling complete 3D environments, our data augmentation approach requires only a few user interactions in combination with 3D models of the target object category. Leveraging our approach, we introduce a novel dataset of augmented urban driving scenes with 360 degree images that are used as environment maps to create realistic lighting and reflections on rendered objects. We analyze the significance of realistic object placement by comparing manual placement by humans to automatic methods based on semantic scene analysis. This allows us to create composite images which exhibit both realistic background appearance as well as a large number of complex object arrangements. Through an extensive set of experiments, we conclude the right set of parameters to produce augmented data which can maximally enhance the performance of instance segmentation models. Further, we demonstrate the utility of the proposed approach on training standard deep models for semantic instance segmentation and object detection of cars in outdoor driving scenarios. We test the models trained on our augmented data on the KITTI 2015 dataset, which we have annotated with pixel-accurate ground truth, and on the Cityscapes dataset. Our experiments demonstrate that the models trained on augmented imagery generalize better than those trained on fully synthetic data or models trained on limited amounts of annotated real data.

avg

pdf Project Page [BibTex]

pdf Project Page [BibTex]


no image
Attention please! Enhanced attention control abilities compensate for instructional impairments in multimedia learning

Wirzberger, M., Rey, G. D.

Journal of Computers in Education, 5(2):243-257, Springer Nature, 2018 (article)

Abstract
Learners exposed to multimedia learning contexts have to deal with a variety of visual stimuli, demanding a conducive design of learning material to maintain limitations in attentional resources. Within the current study, effects and constraints arising from two selected impairing features are investigated in more detail within a computer-based learning task on factor analysis. A sample of 53 students received a combination of textual and pictorial elements that explained the topic, while impaired attention was systematically induced in a 2 × 2 factorial between-subjects design by interrupting system-notifications (with vs. without) and seductive text passages (with vs. without). Learners’ ability for controlled attention was assessed with a standardized psychological attention inventory. Approaching the results, learners receiving seductive text passages spent significantly more time on the learning material. In addition, a moderation effect of attention control abilities on the relationship between interruptions and retention performance resulted. Explanations for the obtained findings are discussed referring to mechanisms of compensation, load, and activation.

re

DOI Project Page [BibTex]


no image
The Computational Challenges of Pursuing Multiple Goals: Network Structure of Goal Systems Predicts Human Performance

Reichman, D., Lieder, F., Bourgin, D. D., Talmon, N., Griffiths, T. L.

PsyArXiv, 2018 (article)

re

DOI [BibTex]

DOI [BibTex]


no image
The moderating role of arousal on the seductive detail effect in a multimedia learning setting

Schneider, S., Wirzberger, M., Rey, G. D.

Applied Cognitive Psychology, Wiley, 2018 (article)

Abstract
Arousal has been found to increase learners' attentional resources. In contrast, seductive details (interesting but learning‐irrelevant information) are considered to distract attention away from relevant information and, thus, hinder learning. However, a possibly moderating role of arousal on the seductive detail effect has not been examined yet. In this study, arousal variations were induced via audio files of false heartbeats. In consequence, 100 participants were randomly assigned to a 2 (with or without seductive details) × 2 (lower vs. higher false heart rates) between‐subjects design. Data on learning performance, cognitive load, motivation, heartbeat frequency, and electro‐dermal activity were collected. Results show learning‐inhibiting effects for seductive details and learning‐enhancing effects for higher false heart rates. Cognitive processes mediate both effects. However, the detrimental effect of seductive details was not present when heart rate was higher. Results indicate that the seductive detail effect is moderated by a learner's state of arousal.

re

DOI [BibTex]

DOI [BibTex]


Learning 3D Shape Completion under Weak Supervision
Learning 3D Shape Completion under Weak Supervision

Stutz, D., Geiger, A.

International Journal of Computer Vision (IJCV), 2018, 2018 (article)

Abstract
We address the problem of 3D shape completion from sparse and noisy point clouds, a fundamental problem in computer vision and robotics. Recent approaches are either data-driven or learning-based: Data-driven approaches rely on a shape model whose parameters are optimized to fit the observations; Learning-based approaches, in contrast, avoid the expensive optimization step by learning to directly predict complete shapes from incomplete observations in a fully-supervised setting. However, full supervision is often not available in practice. In this work, we propose a weakly-supervised learning-based approach to 3D shape completion which neither requires slow optimization nor direct supervision. While we also learn a shape prior on synthetic data, we amortize, i.e., learn, maximum likelihood fitting using deep neural networks resulting in efficient shape completion without sacrificing accuracy. On synthetic benchmarks based on ShapeNet and ModelNet as well as on real robotics data from KITTI and Kinect, we demonstrate that the proposed amortized maximum likelihood approach is able to compete with a fully supervised baseline and outperforms the data-driven approach of Engelmann et al., while requiring less supervision and being significantly faster.

avg

pdf Project Page [BibTex]

pdf Project Page [BibTex]


Object Scene Flow
Object Scene Flow

Menze, M., Heipke, C., Geiger, A.

ISPRS Journal of Photogrammetry and Remote Sensing, 2018 (article)

Abstract
This work investigates the estimation of dense three-dimensional motion fields, commonly referred to as scene flow. While great progress has been made in recent years, large displacements and adverse imaging conditions as observed in natural outdoor environments are still very challenging for current approaches to reconstruction and motion estimation. In this paper, we propose a unified random field model which reasons jointly about 3D scene flow as well as the location, shape and motion of vehicles in the observed scene. We formulate the problem as the task of decomposing the scene into a small number of rigidly moving objects sharing the same motion parameters. Thus, our formulation effectively introduces long-range spatial dependencies which commonly employed local rigidity priors are lacking. Our inference algorithm then estimates the association of image segments and object hypotheses together with their three-dimensional shape and motion. We demonstrate the potential of the proposed approach by introducing a novel challenging scene flow benchmark which allows for a thorough comparison of the proposed scene flow approach with respect to various baseline models. In contrast to previous benchmarks, our evaluation is the first to provide stereo and optical flow ground truth for dynamic real-world urban scenes at large scale. Our experiments reveal that rigid motion segmentation can be utilized as an effective regularizer for the scene flow problem, improving upon existing two-frame scene flow methods. At the same time, our method yields plausible object segmentations without requiring an explicitly trained recognition model for a specific object class.

avg

Project Page [BibTex]

Project Page [BibTex]


no image
Rational metareasoning and the plasticity of cognitive control

Lieder, F., Shenhav, A., Musslick, S., Griffiths, T. L.

{PLoS Computational Biology}, 14(4):e1006043, Public Library of Science, 2018 (article)

re

Project Page Project Page [BibTex]

Project Page Project Page [BibTex]


no image
Over-representation of extreme events in decision making reflects rational use of cognitive resources

Lieder, F., Griffiths, T. L., Hsu, M.

Psychological Review, 125(1):1-32, 2018 (article)

re

[BibTex]

[BibTex]

2016


no image
Sustainable effects of simulator-based training on ecological driving

Lüderitz, C., Wirzberger, M., Karrer-Gauß, K.

In Advances in Ergonomic Design of Systems, Products and Processes. Proceedings of the Annual Meeting of the GfA 2015, pages: 463-475, Springer, 2016 (inbook)

Abstract
Simulation-based driver training offers a promising way to teach ecological driving behavior under controlled, comparable conditions. In a study with 23 professional drivers, we tested the effectiveness of such training. The driving behavior of a training group in a simulated drive with and without instructions were compared. Ten weeks later, a repetition drive tested the long-term effect training. Driving data revealed reduced fuel consumption by ecological driving in both the guided and repetition drives. Driving time decreased significantly in the training and did not differ from driving time after 10 weeks. Results did not achieve significance for transfer to test drives in real traffic situations. This may be due to the small sample size and biased data as a result of unusual driving behavior. Finally, recent and promising approaches to support drivers in maintaining eco-driving styles beyond training situations are outlined.

re

DOI [BibTex]

2016


DOI [BibTex]


Probabilistic Duality for Parallel Gibbs Sampling without Graph Coloring
Probabilistic Duality for Parallel Gibbs Sampling without Graph Coloring

Mescheder, L., Nowozin, S., Geiger, A.

Arxiv, 2016 (article)

Abstract
We present a new notion of probabilistic duality for random variables involving mixture distributions. Using this notion, we show how to implement a highly-parallelizable Gibbs sampler for weakly coupled discrete pairwise graphical models with strictly positive factors that requires almost no preprocessing and is easy to implement. Moreover, we show how our method can be combined with blocking to improve mixing. Even though our method leads to inferior mixing times compared to a sequential Gibbs sampler, we argue that our method is still very useful for large dynamic networks, where factors are added and removed on a continuous basis, as it is hard to maintain a graph coloring in this setup. Similarly, our method is useful for parallelizing Gibbs sampling in graphical models that do not allow for graph colorings with a small number of colors such as densely connected graphs.

avg

pdf [BibTex]


no image
One for all?! Simultaneous examination of load-inducing factors for advancing media-related instructional research

Wirzberger, M., Beege, M., Schneider, S., Nebel, S., Rey, G. D.

Computers {\&} Education, 100, pages: 18-31, Elsevier BV, 2016 (article)

Abstract
In multimedia learning settings, limitations in learners' mental resource capacities need to be considered to avoid impairing effects on learning performance. Based on the prominent and often quoted Cognitive Load Theory, this study investigates the potential of a single experimental approach to provide simultaneous and separate measures for the postulated load-inducing factors. Applying a basal letter-learning task related to the process of working memory updating, intrinsic cognitive load (by varying task complexity), extraneous cognitive load (via inducing split-attention demands) and germane cognitive load (by varying the presence of schemata) were manipulated within a 3 × 2 × 2-factorial full repeated-measures design. The performance of a student sample (N = 96) was inspected regarding reaction times and errors in updating and recall steps. Approaching the results with linear mixed models, the effect of complexity gained substantial strength, whereas the other factors received at least partial significant support. Additionally, interactions between two or all load-inducing factors occurred. Despite various open questions, the study comprises a promising step for the empirical investigation of existing construction yards in cognitive load research.

re

DOI [BibTex]

DOI [BibTex]


Map-Based Probabilistic Visual Self-Localization
Map-Based Probabilistic Visual Self-Localization

Brubaker, M. A., Geiger, A., Urtasun, R.

IEEE Trans. on Pattern Analysis and Machine Intelligence (PAMI), 2016 (article)

Abstract
Accurate and efficient self-localization is a critical problem for autonomous systems. This paper describes an affordable solution to vehicle self-localization which uses odometry computed from two video cameras and road maps as the sole inputs. The core of the method is a probabilistic model for which an efficient approximate inference algorithm is derived. The inference algorithm is able to utilize distributed computation in order to meet the real-time requirements of autonomous systems in some instances. Because of the probabilistic nature of the model the method is capable of coping with various sources of uncertainty including noise in the visual odometry and inherent ambiguities in the map (e.g., in a Manhattan world). By exploiting freely available, community developed maps and visual odometry measurements, the proposed method is able to localize a vehicle to 4m on average after 52 seconds of driving on maps which contain more than 2,150km of drivable roads.

avg ps

pdf Project Page [BibTex]

pdf Project Page [BibTex]

2006


no image
Die Effektivität von schriftlichen und graphischen Warnhinweisen auf Zigarettenschachteln

Petersen, L., Lieder, F.

Zeitschrift f{\"u}r Sozialpsychologie, 37(4):245-258, Verlag Hans Huber, 2006 (article)

re

[BibTex]

2006


[BibTex]