Judith Bütepage

Amir Hossein Akhavan Rahnama, Judith Butepage

Local explanations of learning-to-rank (LTR) models are thought to extract the most important features that contribute to the ranking predicted by the LTR model for a single data point. Evaluating the accuracy of such explanations is challenging since the ground truth feature importance scores are not available for most modern LTR models. In this work, we propose a systematic evaluation technique for explanations of LTR models. Instead of using black-box models, such as neural networks, we propose to focus on tree-based LTR models, from which we can extract the ground truth feature importance scores using decision paths. Once extracted, we can directly compare the ground truth feature importance scores to the feature importance scores generated with explanation techniques. We compare two recently proposed explanation techniques for LTR models when using decision trees and gradient boosting models on the MQ2008 dataset. We show that the explanation accuracy in these techniques can largely vary depending on the explained model and even which data point is explained.

Zhen Han, Mattias Teye, Derek Yadgaroff et al.

The training of high-quality, robust machine learning models for speech-driven 3D facial animation requires a large, diverse dataset of high-quality audio-animation pairs. To overcome the lack of such a dataset, recent work has introduced large pre-trained speech encoders that are robust to variations in the input audio and, therefore, enable the facial animation model to generalize across speakers, audio quality, and languages. However, the resulting facial animation models are prohibitively large and lend themselves only to offline inference on a dedicated machine. In this work, we explore on-device, real-time facial animation models in the context of game development. We overcome the lack of large datasets by using hybrid knowledge distillation with pseudo-labeling. Given a large audio dataset, we employ a high-performing teacher model to train very small student models. In contrast to the pre-trained speech encoders, our student models only consist of convolutional and fully-connected layers, removing the need for attention context or recurrent updates. In our experiments, we demonstrate that we can reduce the memory footprint to up to 3.4 MB and required future audio context to up to 81 ms while maintaining high-quality animations. This paves the way for on-device inference, an important step towards realistic, model-driven digital characters.

Amir Hossein Akhavan Rahnama, Judith Butepage, Pierre Geurts et al.

Evaluating explanation techniques using human subjects is costly, time-consuming and can lead to subjectivity in the assessments. To evaluate the accuracy of local explanations, we require access to the true feature importance scores for a given instance. However, the prediction function of a model usually does not decompose into linear additive terms that indicate how much a feature contributes to the output. In this work, we suggest to instead focus on the log odds ratio (LOR) of the prediction function, which naturally decomposes into additive terms for logistic regression and naive Bayes. We demonstrate how we can benchmark different explanation techniques in terms of their similarity to the LOR scores based on our proposed approach. In the experiments, we compare prominent local explanation techniques and find that the performance of the techniques can depend on the underlying model, the dataset, which data point is explained, the normalization of the data and the similarity metric.

Thomas Ziegler, Judith Butepage, Michael C. Welle et al.

Research on automated, image based identification of clothing categories and fashion landmarks has recently gained significant interest due to its potential impact on areas such as robotic clothing manipulation, automated clothes sorting and recycling, and online shopping. Several public and annotated fashion datasets have been created to facilitate research advances in this direction. In this work, we make the first step towards leveraging the data and techniques developed for fashion image analysis in vision-based robotic clothing manipulation tasks. We focus on techniques that can generalize from large-scale fashion datasets to less structured, small datasets collected in a robotic lab. Specifically, we propose training data augmentation methods such as elastic warping, and model adjustments such as rotation invariant convolutions to make the model generalize better. Our experiments demonstrate that our approach outperforms stateof-the art models with respect to clothing category classification and fashion landmark detection when tested on previously unseen datasets. Furthermore, we present experimental results on a new dataset composed of images where a robot holds different garments, collected in our lab.

Judith Bütepage, Hedvig Kjellström, Danica Kragic

Human behavior is a continuous stochastic spatio-temporal process which is governed by semantic actions and affordances as well as latent factors. Therefore, video-based human activity modeling is concerned with a number of tasks such as inferring current and future semantic labels, predicting future continuous observations as well as imagining possible future label and feature sequences. In this paper we present a semi-supervised probabilistic deep latent variable model that can represent both discrete labels and continuous observations as well as latent dynamics over time. This allows the model to solve several tasks at once without explicit fine-tuning. We focus here on the tasks of action classification, detection, prediction and anticipation as well as motion prediction and synthesis based on 3D human activity data recorded with Kinect. We further extend the model to capture hierarchical label structure and to model the dependencies between multiple entities, such as a human and objects. Our experiments demonstrate that our principled approach to human activity modeling can be used to detect current and anticipate future semantic labels and to predict and synthesize future label and feature sequences. When comparing our model to state-of-the-art approaches, which are specifically designed for e.g. action classification, we find that our probabilistic formulation outperforms or is comparable to these task specific models.

Judith Bütepage, Danica Kragic

Successful Human-Robot collaboration requires a predictive model of human behavior. The robot needs to be able to recognize current goals and actions and to predict future activities in a given context. However, the spatio-temporal sequence of human actions is difficult to model since latent factors such as intention, task, knowledge, intuition and preference determine the action choices of each individual. In this work we introduce semi-supervised variational recurrent neural networks which are able to a) model temporal distributions over latent factors and the observable feature space, b) incorporate discrete labels such as activity type when available, and c) generate possible future action sequences on both feature and label level. We evaluate our model on the Cornell Activity Dataset CAD-120 dataset. Our model outperforms state-of-the-art approaches in both activity and affordance detection and anticipation. Additionally, we show how samples of possible future action sequences are in line with past observations.

Cheng Zhang, Judith Butepage, Hedvig Kjellstrom et al.

Many modern unsupervised or semi-supervised machine learning algorithms rely on Bayesian probabilistic models. These models are usually intractable and thus require approximate inference. Variational inference (VI) lets us approximate a high-dimensional Bayesian posterior with a simpler variational distribution by solving an optimization problem. This approach has been successfully used in various models and large-scale applications. In this review, we give an overview of recent trends in variational inference. We first introduce standard mean field variational inference, then review recent advances focusing on the following aspects: (a) scalable VI, which includes stochastic approximations, (b) generic VI, which extends the applicability of VI to a large class of otherwise intractable models, such as non-conjugate models, (c) accurate VI, which includes variational models beyond the mean field approximation or with atypical divergences, and (d) amortized VI, which implements the inference over local latent variables with inference networks. Finally, we provide a summary of promising future research directions.

Judith Bütepage, Hedvig Kjellström, Danica Kragic

Fluent and safe interactions of humans and robots require both partners to anticipate the others' actions. A common approach to human intention inference is to model specific trajectories towards known goals with supervised classifiers. However, these approaches do not take possible future movements into account nor do they make use of kinematic cues, such as legible and predictable motion. The bottleneck of these methods is the lack of an accurate model of general human motion. In this work, we present a conditional variational autoencoder that is trained to predict a window of future human motion given a window of past frames. Using skeletal data obtained from RGB depth images, we show how this unsupervised approach can be used for online motion prediction for up to 1660 ms. Additionally, we demonstrate online target prediction within the first 300-500 ms after motion onset without the use of target specific training data. The advantage of our probabilistic approach is the possibility to draw samples of possible future motions. Finally, we investigate how movements and kinematic cues are represented on the learned low dimensional manifold.

Judith Bütepage, Michael Black, Danica Kragic et al.

Generative models of 3D human motion are often restricted to a small number of activities and can therefore not generalize well to novel movements or applications. In this work we propose a deep learning framework for human motion capture data that learns a generic representation from a large corpus of motion capture data and generalizes well to new, unseen, motions. Using an encoding-decoding network that learns to predict future 3D poses from the most recent past, we extract a feature representation of human motion. Most work on deep learning for sequence prediction focuses on video and speech. Since skeletal data has a different structure, we present and evaluate different network architectures that make different assumptions about time dependencies and limb correlations. To quantify the learned features, we use the output of different layers for action classification and visualize the receptive fields of the network units. Our method outperforms the recent state of the art in skeletal motion prediction even though these use action specific training data. Our results show that deep feedforward networks, trained from a generic mocap database, can successfully be used for feature extraction from human motion data and that this representation can be used as a foundation for classification and prediction.

Ali Ghadirzadeh, Judith Bütepage, Atsuto Maki et al.

Modeling of physical human-robot collaborations is generally a challenging problem due to the unpredictive nature of human behavior. To address this issue, we present a data-efficient reinforcement learning framework which enables a robot to learn how to collaborate with a human partner. The robot learns the task from its own sensorimotor experiences in an unsupervised manner. The uncertainty of the human actions is modeled using Gaussian processes (GP) to implement action-value functions. Optimal action selection given the uncertain GP model is ensured by Bayesian optimization. We apply the framework to a scenario in which a human and a PR2 robot jointly control the ball position on a plank based on vision and force/torque data. Our experimental results show the suitability of the proposed method in terms of fast and data-efficient model learning, optimal action selection under uncertainties and equal role sharing between the partners.

Ali Ghadirzadeh, Judith Bütepage, Danica Kragic et al.

We present a general framework to autonomously achieve a task, where autonomy is acquired by learning sensorimotor patterns of a robot, while it is interacting with its environment. To accomplish the task, using the learned sensorimotor contingencies, our approach predicts a sequence of actions that will lead to the desirable observations. Gaussian processes (GP) with automatic relevance determination is used to learn the sensorimotor mapping. In this way, relevant sensory and motor components can be systematically found in high-dimensional sensory and motor spaces. We propose an incremental GP learning strategy, which discerns between situations, when an update or an adaptation must be implemented. RRT* is exploited to enable long-term planning and generating a sequence of states that lead to a given goal; while a gradient-based search finds the optimum action to steer to a neighbouring state in a single time step. Our experimental results prove the successfulness of the proposed framework to learn a joint space controller with high data dimensions (10$\times$15). It demonstrates short training phase (less than 12 seconds), real-time performance and rapid adaptations capabilities.