Author name cluster

Thomas Brox

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

42 papers

2 author rows

ICLR Conference 2025 Conference Paper

Bridging Information Asymmetry in Text-video Retrieval: A Data-centric Approach

Zechen Bai
Tianjun Xiao
Tong He 0002
Pichao Wang
Zheng Zhang 0001
Thomas Brox
Mike Zheng Shou

As online video content rapidly grows, the task of text-video retrieval (TVR) becomes increasingly important. A key challenge in TVR is the information asymmetry between video and text: videos are inherently richer in information, while their textual descriptions often capture only fragments of this complexity. This paper introduces a novel, data-centric framework to bridge this gap by enriching textual representations to better match the richness of video content. During training, videos are segmented into event-level clips and captioned to ensure comprehensive coverage. During retrieval, a large language model (LLM) generates semantically diverse queries to capture a broader range of possible matches. To enhance retrieval efficiency, we propose a query selection mechanism that identifies the most relevant and diverse queries, reducing computational cost while improving accuracy. Our method achieves state-of-the-art results across multiple benchmarks, demonstrating the power of data-centric approaches in addressing information asymmetry in TVR. This work paves the way for new research focused on leveraging data to improve cross-modal retrieval.

TMLR Journal 2025 Journal Article

Constrained Reinforcement Learning with Smoothed Log Barrier Function

Baohe Zhang
Yuan Zhang
Hao Zhu
Shengchao Yan
Thomas Brox
Joschka Boedecker

Deploying reinforcement learning (RL) in real-world systems often requires satisfying strict safety constraints during both training and deployment, which simple reward shaping typically fails to enforce. Existing constrained RL algorithms frequently face several major challenges, including instabilities during training and overly conservative policies. To overcome these limitations, we propose CSAC-LB (Constrained Soft Actor-Critic with Log Barrier), a model-free, sample-efficient, off-policy algorithm that requires no pre-training. CSAC-LB integrates a linear smoothed log barrier function into the actor’s objective, providing a numerically stable, non-vanishing gradient that enables the agent to quickly recover from unsafe states while avoiding the instability of traditional interior-point methods. To further enhance safety and mitigate the underestimation of constraint violations, we employ a pessimistic double-critic architecture for the cost function, taking the maximum of two cost Q-networks to conservatively guide the policy. Through extensive experiments on challenging constrained control tasks, we demonstrate that CSAC-LB significantly outperforms baselines by consistently achieving high returns while strictly adhering to safety constraints. Our results establish CSAC-LB as a robust and stable solution for applying RL to safety-critical domains.

IROS Conference 2025 Conference Paper

Label-Efficient LiDAR Semantic Segmentation with 2D-3D Vision Transformer Adapters

Julia Hindel
Rohit Mohan
Jelena Bratulic
Daniele Cattaneo 0001
Thomas Brox
Abhinav Valada

LiDAR semantic segmentation models are typically trained from random initialization as universal pre-training is hindered by the lack of large, diverse datasets. Moreover, most point cloud segmentation architectures incorporate custom network layers, limiting the transferability of advances from vision-based architectures. Inspired by recent advances in universal foundation models, we propose BALViT, a novel approach that leverages frozen vision models as amodal feature encoders for learning strong LiDAR encoders. Specifically, BALViT incorporates both range-view and bird’s-eye-view LiDAR encoding mechanisms, which we combine through a novel 2D-3D adapter. While the range-view features are processed through a frozen image backbone, our bird’s-eye-view branch enhances them through multiple cross-attention interactions. Thereby, we continuously improve the vision network with domain-dependent knowledge, resulting in a strong label-efficient LiDAR encoding mechanism. Extensive evaluations of BALViT on the SemanticKITTI and nuScenes benchmarks demonstrate that it outperforms state-of-the-art methods on small data regimes. We make the code and models publicly available at http://balvit.cs.uni-freiburg.de.

NeurIPS Conference 2025 Conference Paper

Overcoming Challenges of Long-Horizon Prediction in Driving World Models

Arian Mousakhan
Sudhanshu Mittal
Silvio Galesso
Karim Farid
Thomas Brox

Existing world models for autonomous driving struggle with long-horizon generation and generalization to challenging scenarios. In this work, we develop a model using simple design choices, and without additional supervision or sensors, such as maps, depth, or multiple cameras. We show that our model yields state-of-the-art performance, despite having only 469M parameters and being trained on 280h of video data. It particularly stands out in difficult scenarios like turning maneuvers and urban traffic. We test whether discrete token models possibly have advantages over continuous models based on flow matching. To this end, we set up a hybrid tokenizer that is compatible with both approaches and allows for a side-by-side comparison. Our study concludes in favor of the continuous autoregressive model, which is less brittle on individual design choices and more powerful than the model built on discrete tokens. Project page with code, model checkpoints and visualization can be found here: https: //lmb-freiburg. github. io/orbis. github. io/

TMLR Journal 2025 Journal Article

Selective Concept Bottleneck Models Without Predefined Concepts

Simon Schrodi
Julian Schur
Max Argus
Thomas Brox

Concept-based models like Concept Bottleneck Models (CBMs) have garnered significant interest for improving model interpretability by first predicting human-understandable concepts before mapping them to the output classes. Early approaches required costly concept annotations. To alleviate this, recent methods utilized large language models to automatically generate class-specific concept descriptions and learned mappings from a pretrained black-box model’s raw features to these concepts using vision-language models. However, these approaches assume prior knowledge of which concepts the black-box model has learned. In this work, we discover the concepts encoded by the model through unsupervised concept discovery techniques instead. We further leverage a simple input-dependent concept selection mechanism that dynamically retains a sparse set of relevant concepts of each input, enhancing both sparsity and interpretability. Our approach not only improves downstream performance, but also needs significantly fewer concepts for accurate classification. Lastly, we show how large vision-language models can guide the editing of our models' weights to correct model errors.

ICLR Conference 2025 Conference Paper

Two Effects, One Trigger: On the Modality Gap, Object Bias, and Information Imbalance in Contrastive Vision-Language Models

Simon Schrodi
David T. Hoffmann
Max Argus
Volker Fischer 0003
Thomas Brox

Contrastive vision-language models (VLMs), like CLIP, have gained popularity for their versatile applicability to various downstream tasks. Despite their successes in some tasks, like zero-shot object recognition, they perform surprisingly poor on other tasks, like attribute recognition. Previous work has attributed these challenges to the modality gap, a separation of image and text in the shared representation space, and to a bias towards objects over other factors, such as attributes. In this analysis paper, we investigate both phenomena thoroughly. We evaluated off-the-shelf VLMs and while the gap's influence on performance is typically overshadowed by other factors, we find indications that closing the gap indeed leads to improvements. Moreover, we find that, contrary to intuition, only few embedding dimensions drive the gap and that the embedding spaces are differently organized. To allow for a clean study of object bias, we introduce a definition and a corresponding measure of it. Equipped with this tool, we find that object bias does not lead to worse performance on other concepts, such as attributes per se. However, why do both phenomena, modality gap and object bias, emerge in the first place? To answer this fundamental question and uncover some of the inner workings of contrastive VLMs, we conducted experiments that allowed us to control the amount of shared information between the modalities. These experiments revealed that the driving factor behind both the modality gap and the object bias, is an information imbalance between images and captions, and unveiled an intriguing connection between the modality gap and entropy of the logits.

ICML Conference 2025 Conference Paper

When and How Does CLIP Enable Domain and Compositional Generalization?

Elias Kempf
Simon Schrodi
Max Argus
Thomas Brox

The remarkable generalization performance of contrastive vision-language models like CLIP is often attributed to the diversity of their training distributions. However, key questions remain unanswered: Can CLIP generalize to an entirely unseen domain when trained on a diverse mixture of domains (domain generalization)? Can it generalize to unseen classes within partially seen domains (compositional generalization)? What factors affect such generalization? To answer these questions, we trained CLIP models on systematically constructed training distributions with controlled domain diversity and object class exposure. Our experiments show that domain diversity is essential for both domain and compositional generalization, yet compositional generalization can be surprisingly weaker than domain generalization when the training distribution contains a suboptimal subset of the test domain. Through data-centric and mechanistic analyses, we find that successful generalization requires the learning of sufficiently shared representations in intermediate layers and circuits.

ICRA Conference 2024 Conference Paper

Amodal Optical Flow

Maximilian Luz
Rohit Mohan
Ahmed Rida Sekkat
Oliver Sawade
Elmar Matthes
Thomas Brox
Abhinav Valada

Optical flow estimation is very challenging in situations with transparent or occluded objects. In this work, we address these challenges at the task level by introducing Amodal Optical Flow, which integrates optical flow with amodal perception. Instead of only representing the visible regions, we define amodal optical flow as a multi-layered pixel-level motion field that encompasses both visible and occluded regions of the scene. To facilitate research on this new task, we extend the AmodalSynthDrive dataset to include pixel-level labels for amodal optical flow estimation. We present several strong baselines, along with the Amodal Flow Quality metric to quantify the performance in an interpretable manner. Furthermore, we propose the novel AmodalFlowNet as an initial step toward addressing this task. AmodalFlowNet consists of a transformer-based cost-volume encoder paired with a recurrent transformer decoder which facilitates recurrent hierarchical feature propagation and amodal semantic grounding. We demonstrate the tractability of amodal optical flow in extensive experiments and show its utility for downstream tasks such as panoptic tracking. We make the dataset, code, and trained models publicly available at http://amodal-flow. cs. uni-freiburg. de.

ICRA Conference 2024 Conference Paper

Compositional Servoing by Recombining Demonstrations

Max Argus
Abhijeet Nayak
Martin Büchner
Silvio Galesso
Abhinav Valada
Thomas Brox

Learning-based manipulation policies from image inputs often show weak task transfer capabilities. In contrast, visual servoing methods allow efficient task transfer in high-precision scenarios while requiring only a few demonstrations. In this work, we present a framework that formulates the visual servoing task as graph traversal. Our method not only extends the robustness of visual servoing, but also enables multitask capability based on a few task-specific demonstrations. We construct demonstration graphs by splitting existing demonstrations and recombining them. In order to traverse the demonstration graph in the inference case, we utilize a similarity function that helps select the best demonstration for a specific task. This enables us to compute the shortest path through the graph. Ultimately, we show that recombining demonstrations leads to higher task-respective success. We present extensive simulation and real-world experimental results that demonstrate the efficacy of our approach.

ICLR Conference 2024 Conference Paper

CrossQ: Batch Normalization in Deep Reinforcement Learning for Greater Sample Efficiency and Simplicity

Aditya Bhatt 0001
Daniel Palenicek
Boris Belousov
Max Argus
Artemij Amiranashvili
Thomas Brox
Jan Peters 0001

Sample efficiency is a crucial problem in deep reinforcement learning. Recent algorithms, such as REDQ and DroQ, found a way to improve the sample efficiency by increasing the update-to-data (UTD) ratio to 20 gradient update steps on the critic per environment sample. However, this comes at the expense of a greatly increased computational cost. To reduce this computational burden, we introduce CrossQ: A lightweight algorithm for continuous control tasks that makes careful use of Batch Normalization and removes target networks to surpass the current state-of-the-art in sample efficiency while maintaining a low UTD ratio of 1. Notably, CrossQ does not rely on advanced bias-reduction schemes used in current methods. CrossQ's contributions are threefold: (1) it matches or surpasses current state-of-the-art methods in terms of sample efficiency, (2) it substantially reduces the computational cost compared to REDQ and DroQ, (3) it is easy to implement, requiring just a few lines of code on top of SAC.

EWRL Workshop 2024 Workshop Paper

CrossQ: Batch Normalization in Deep Reinforcement Learning for Greater Sample Efficiency and Simplicity

Aditya Bhatt
Daniel Palenicek
Boris Belousov
Max Argus
Artemij Amiranashvili
Thomas Brox
Jan Peters

Sample efficiency is a crucial problem in deep reinforcement learning. Recent algorithms, such as REDQ and DroQ, found a way to improve the sample efficiency by increasing the update-to-data (UTD) ratio to 20 gradient update steps on the critic per environment sample. However, this comes at the expense of a greatly increased computational cost. To reduce this computational burden, we introduce CrossQ: A lightweight algorithm for continuous control tasks that makes careful use of Batch Normalization and removes target networks to surpass the current state-of-the-art in sample efficiency while maintaining a low UTD ratio of 1. Notably, CrossQ does not rely on advanced bias-reduction schemes used in current methods. CrossQ's contributions are threefold: (1) it matches or surpasses current state-of-the-art methods in terms of sample efficiency, (2) it substantially reduces the computational cost compared to REDQ and DroQ, (3) it is easy to implement, requiring just a few lines of code on top of SAC.

IROS Conference 2024 Conference Paper

DITTO: Demonstration Imitation by Trajectory Transformation

Nick Heppert
Max Argus
Tim Welschehold
Thomas Brox
Abhinav Valada

Teaching robots new skills quickly and conveniently is crucial for the broader adoption of robotic systems. In this work, we address the problem of one-shot imitation from a single human demonstration, given by an RGB-D video recording. We propose a two-stage process. In the first stage we extract the demonstration trajectory offline. This entails segmenting manipulated objects and determining their relative motion in relation to secondary objects such as containers. In the online trajectory generation stage, we first re-detect all objects, then warp the demonstration trajectory to the current scene and execute it on the robot. To complete these steps, our method leverages several ancillary models, including those for segmentation, relative object pose estimation, and grasp prediction. We systematically evaluate different combinations of correspondence and re-detection methods to validate our design decision across a diverse range of tasks. Specifically, we collect and quantitatively test on demonstrations of ten different tasks including pick-and-place tasks as well as articulated object manipulation. Finally, we perform extensive evaluations on a real robot system to demonstrate the effectiveness and utility of our approach in real-world scenarios. We make the code publicly available at http://ditto.cs.uni-freiburg.de.

ICML Conference 2024 Conference Paper

Eureka-Moments in Transformers: Multi-Step Tasks Reveal Softmax Induced Optimization Problems

David T. Hoffmann
Simon Schrodi
Jelena Bratulic
Nadine Behrmann
Volker Fischer 0003
Thomas Brox

In this work, we study rapid improvements of the training loss in transformers when being confronted with multi-step decision tasks. We found that transformers struggle to learn the intermediate task and both training and validation loss saturate for hundreds of epochs. When transformers finally learn the intermediate task, they do this rapidly and unexpectedly. We call these abrupt improvements Eureka-moments, since the transformer appears to suddenly learn a previously incomprehensible concept. We designed synthetic tasks to study the problem in detail, but the leaps in performance can be observed also for language modeling and in-context learning (ICL). We suspect that these abrupt transitions are caused by the multi-step nature of these tasks. Indeed, we find connections and show that ways to improve on the synthetic multi-step tasks can be used to improve the training of language modeling and ICL. Using the synthetic data we trace the problem back to the Softmax function in the self-attention block of transformers and show ways to alleviate the problem. These fixes reduce the required number of training steps, lead to higher likelihood to learn the intermediate task, to higher final accuracy and training becomes more robust to hyper-parameters.

ICLR Conference 2024 Conference Paper

Open-ended VQA benchmarking of Vision-Language models by exploiting Classification datasets and their semantic hierarchy

Simon Ging
María Alejandra Bravo
Thomas Brox

The evaluation of text-generative vision-language models is a challenging yet crucial endeavor. By addressing the limitations of existing Visual Question Answering (VQA) benchmarks and proposing innovative evaluation methodologies, our research seeks to advance our understanding of these models’ capabilities. We propose a novel VQA benchmark based on well-known visual classification datasets which allows a granular evaluation of text-generative vision-language models and their comparison with discriminative vision-language models. To improve the assessment of coarse answers on fine-grained classification tasks, we suggest using the semantic hierarchy of the label space to ask automatically generated follow-up questions about the ground-truth category. Finally, we compare traditional NLP and LLM-based metrics for the problem of evaluating model predictions given ground-truth answers. We perform a human evaluation study upon which we base our decision on the final metric. We apply our benchmark to a suite of vision-language models and show a detailed comparison of their abilities on object, action, and attribute classification. Our contributions aim to lay the foundation for more precise and meaningful assessments, facilitating targeted progress in the exciting field of vision-language modeling.

ICLR Conference 2023 Conference Paper

Bridging the Gap to Real-World Object-Centric Learning

Maximilian Seitzer
Max Horn
Andrii Zadaianchuk
Dominik Zietlow
Tianjun Xiao
Carl-Johann Simon-Gabriel
Tong He 0002
Zheng Zhang 0001

Humans naturally decompose their environment into entities at the appropriate level of abstraction to act in the world. Allowing machine learning algorithms to derive this decomposition in an unsupervised way has become an important line of research. However, current methods are restricted to simulated data or require additional information in the form of motion or depth in order to successfully discover objects. In this work, we overcome this limitation by showing that reconstructing features from models trained in a self-supervised manner is a sufficient training signal for object-centric representations to arise in a fully unsupervised way. Our approach, DINOSAUR, significantly out-performs existing object-centric learning models on simulated data and is the first unsupervised object-centric model that scales to real world-datasets such as COCO and PASCAL VOC. DINOSAUR is conceptually simple and shows competitive performance compared to more involved pipelines from the computer vision literature.

NeurIPS Conference 2023 Conference Paper

Construction of Hierarchical Neural Architecture Search Spaces based on Context-free Grammars

Simon Schrodi
Danny Stoll
Binxin Ru
Rhea Sukthanker
Thomas Brox
Frank Hutter

The discovery of neural architectures from simple building blocks is a long-standing goal of Neural Architecture Search (NAS). Hierarchical search spaces are a promising step towards this goal but lack a unifying search space design framework and typically only search over some limited aspect of architectures. In this work, we introduce a unifying search space design framework based on context-free grammars that can naturally and compactly generate expressive hierarchical search spaces that are 100s of orders of magnitude larger than common spaces from the literature. By enhancing and using their properties, we effectively enable search over the complete architecture and can foster regularity. Further, we propose an efficient hierarchical kernel design for a Bayesian Optimization search strategy to efficiently search over such huge spaces. We demonstrate the versatility of our search space design framework and show that our search strategy can be superior to existing NAS approaches. Code is available at https: //github. com/automl/hierarchical nas construction.

ICLR Conference 2023 Conference Paper

Unsupervised Semantic Segmentation with Self-supervised Object-centric Representations

Andrii Zadaianchuk
Matthäus Kleindessner
Yi Zhu
Francesco Locatello
Thomas Brox

In this paper, we show that recent advances in self-supervised representation learning enable unsupervised object discovery and semantic segmentation with a performance that matches the state of the field on supervised semantic segmentation 10 years ago. We propose a methodology based on unsupervised saliency masks and self-supervised feature clustering to kickstart object discovery followed by training a semantic segmentation network on pseudo-labels to bootstrap the system on images with multiple objects. We show that while being conceptually simple our proposed baseline is surprisingly strong. We present results on PASCAL VOC that go far beyond the current state of the art (50.0 mIoU), and we report for the first time results on MS COCO for the whole set of 81 classes: our method discovers 34 categories with more than 20% IoU, while obtaining an average IoU of 19.6 for all 81 categories.

NeurIPS Conference 2022 Conference Paper

Assaying Out-Of-Distribution Generalization in Transfer Learning

Florian Wenzel
Andrea Dittadi
Peter Gehler
Carl-Johann Simon-Gabriel
Max Horn
Dominik Zietlow
David Kernert
Chris Russell

Since out-of-distribution generalization is a generally ill-posed problem, various proxy targets (e. g. , calibration, adversarial robustness, algorithmic corruptions, invariance across shifts) were studied across different research programs resulting in different recommendations. While sharing the same aspirational goal, these approaches have never been tested under the same experimental conditions on real data. In this paper, we take a unified view of previous work, highlighting message discrepancies that we address empirically, and providing recommendations on how to measure the robustness of a model and how to improve it. To this end, we collect 172 publicly available dataset pairs for training and out-of-distribution evaluation of accuracy, calibration error, adversarial attacks, environment invariance, and synthetic corruptions. We fine-tune over 31k networks, from nine different architectures in the many- and few-shot setting. Our findings confirm that in- and out-of-distribution accuracies tend to increase jointly, but show that their relation is largely dataset-dependent, and in general more nuanced and more complex than posited by previous, smaller scale studies.

IROS Conference 2022 Conference Paper

Conditional Visual Servoing for Multi-Step Tasks

Sergio Izquierdo
Max Argus
Thomas Brox

Visual Servoing has been effectively used to move a robot into specific target locations or to track a recorded demonstration. It does not require manual programming, but it is typically limited to settings where one demonstration maps to one environment state. We propose a modular approach to extend visual servoing to scenarios with multiple demonstration sequences. We call this conditional servoing, as we choose the next demonstration conditioned on the observation of the robot. This method presents an appealing strategy to tackle multi-step problems, as individual demonstrations can be combined flexibly into a control policy. We propose different selection functions and compare them on a shape-sorting task in simulation. With the reprojection error yielding the best overall results, we implement this selection function on a real robot and show the efficacy of the proposed conditional servoing. For videos of our experiments, please check out our project page: https://lmb.informatik.uni-freiburg.de/projects/conditional_servoing/

AAAI Conference 2022 Conference Paper

Ranking Info Noise Contrastive Estimation: Boosting Contrastive Learning via Ranked Positives

David T. Hoffmann
Nadine Behrmann
Juergen Gall
Thomas Brox
Mehdi Noroozi

This paper introduces Ranking Info Noise Contrastive Estimation (RINCE), a new member in the family of InfoNCE losses that preserves a ranked ordering of positive samples. In contrast to the standard InfoNCE loss, which requires a strict binary separation of the training pairs into similar and dissimilar samples, RINCE can exploit information about a similarity ranking for learning a corresponding embedding space. We show that the proposed loss function learns favorable embeddings compared to the standard InfoNCE whenever at least noisy ranking information can be obtained or when the definition of positives and negatives is blurry. We demonstrate this for a supervised classification task with additional superclass labels and noisy similarity scores. Furthermore, we show that RINCE can also be applied to unsupervised training with experiments on unsupervised representation learning from videos. In particular, the embedding yields higher classification accuracy, retrieval rates and performs better in out-of-distribution detection than the standard InfoNCE loss.

ICLR Conference 2022 Conference Paper

You Mostly Walk Alone: Analyzing Feature Attribution in Trajectory Prediction

Osama Makansi
Julius von Kügelgen
Francesco Locatello
Peter V. Gehler
Dominik Janzing
Thomas Brox
Bernhard Schölkopf

Predicting the future trajectory of a moving agent can be easy when the past trajectory continues smoothly but is challenging when complex interactions with other agents are involved. Recent deep learning approaches for trajectory prediction show promising performance and partially attribute this to successful reasoning about agent-agent interactions. However, it remains unclear which features such black-box models actually learn to use for making predictions. This paper proposes a procedure that quantifies the contributions of different cues to model performance based on a variant of Shapley values. Applying this procedure to state-of-the-art trajectory prediction methods on standard benchmark datasets shows that they are, in fact, unable to reason about interactions. Instead, the past trajectory of the target is the only feature used for predicting its future. For a task with richer social interaction patterns, on the other hand, the tested models do pick up such interactions to a certain extent, as quantified by our feature attribution method. We discuss the limits of the proposed method and its links to causality.

AAAI Conference 2021 Conference Paper

Explicitly Modeled Attention Maps for Image Classification

Andong Tan
Duc Tam Nguyen
Maximilian Dax
Matthias Nießner
Thomas Brox

Self-attention networks have shown remarkable progress in computer vision tasks such as image classification. The main benefit of the self-attention mechanism is the ability to capture long-range feature interactions in attention-maps. However, the computation of attention-maps requires a learnable key, query, and positional encoding, whose usage is often not intuitive and computationally expensive. To mitigate this problem, we propose a novel self-attention module with explicitly modeled attention-maps using only a single learnable parameter for low computational overhead. The design of explicitly modeled attention-maps using geometric prior is based on the observation that the spatial context for a given pixel within an image is mostly dominated by its neighbors, while more distant pixels have a minor contribution. Concretely, the attention-maps are parametrized via simple functions (e. g. , Gaussian kernel) with a learnable radius, which is modeled independently of the input content. Our evaluation shows that our method achieves an accuracy improvement of up to 2. 2% over the ResNet-baselines in ImageNet ILSVRC and outperforms other self-attention methods such as AA-ResNet152 in accuracy by 0. 9% with 6. 4% fewer parameters and 6. 7% fewer GFLOPs. This result empirically indicates the value of incorporating geometric prior into self-attention mechanism when applied in image classification.

ICRA Conference 2020 Conference Paper

Adaptive Curriculum Generation from Demonstrations for Sim-to-Real Visuomotor Control

Lukás Hermann
Max Argus
Andreas Eitel
Artemij Amiranashvili
Wolfram Burgard
Thomas Brox

We propose Adaptive Curriculum Generation from Demonstrations (ACGD) for reinforcement learning in the presence of sparse rewards. Rather than designing shaped reward functions, ACGD adaptively sets the appropriate task difficulty for the learner by controlling where to sample from the demonstration trajectories and which set of simulation parameters to use. We show that training vision-based control policies in simulation while gradually increasing the difficulty of the task via ACGD improves the policy transfer to the real world. The degree of domain randomization is also gradually increased through the task difficulty. We demonstrate zero-shot transfer for two real-world manipulation tasks: pick-and-stow and block stacking. A video showing the results can be found at https://lmb.informatik.uni-freiburg.de/projects/curriculum/.

NeurIPS Conference 2020 Conference Paper

COOT: Cooperative Hierarchical Transformer for Video-Text Representation Learning

Simon Ging
Mohammadreza Zolfaghari
Hamed Pirsiavash
Thomas Brox

Many real-world video-text tasks involve different levels of granularity, such as frames and words, clip and sentences or videos and paragraphs, each with distinct semantics. In this paper, we propose a Cooperative hierarchical Transformer (COOT) to leverage this hierarchy information and model the interactions between different levels of granularity and different modalities. The method consists of three major components: an attention-aware feature aggregation layer, which leverages the local temporal context (intra-level, e. g. , within a clip), a contextual transformer to learn the interactions between low-level and high-level semantics (inter-level, e. g. clip-video, sentence-paragraph), and a cross-modal cycle-consistency loss to connect video and text. The resulting method compares favorably to the state of the art on several benchmarks while having few parameters.

IROS Conference 2020 Conference Paper

FlowControl: Optical Flow Based Visual Servoing

Max Argus
Lukás Hermann
Jon Long
Thomas Brox

One-shot imitation is the vision of robot programming from a single demonstration, rather than by tedious construction of computer code. We present a practical method for realizing one-shot imitation for manipulation tasks, exploiting modern learning-based optical flow to perform real-time visual servoing. Our approach, which we call FlowControl, continuously tracks a demonstration video, using a specified foreground mask to attend to an object of interest. Using RGB-D observations, FlowControl requires no 3D object models, and is easy to set up. FlowControl inherits great robustness to visual appearance from decades of work in optical flow. We exhibit FlowControl on a range of problems, including ones requiring very precise motions, and ones requiring the ability to generalize.

IROS Conference 2020 Conference Paper

Improving Unimodal Object Recognition with Multimodal Contrastive Learning

Johannes Meyer
Andreas Eitel
Thomas Brox
Wolfram Burgard

Robots perceive their environment using various sensor modalities, e. g. , vision, depth, sound or touch. Each modality provides complementary information for perception. However, while it can be assumed that all modalities are available for training, when deploying the robot in real-world scenarios the sensor setup often varies. In order to gain flexibility with respect to the deployed sensor setup we propose a new multimodal approach within the framework of contrastive learning. In particular, we consider the case of learning from RGB-D images while testing with one modality available, i. e. , exclusively RGB or depth. We leverage contrastive learning to capture high-level information between different modalities in a compact feature embedding. We extensively evaluate our multimodal contrastive learning method on the Falling Things dataset and learn representations that outperform prior methods for RGB-D object recognition on the NYU-D dataset. Our code and details on the used datasets are available at: https://github.com/meyerjo/MultiModalContrastiveLearning.

ICLR Conference 2020 Conference Paper

SELF: Learning to Filter Noisy Labels with Self-Ensembling

Duc Tam Nguyen
Chaithanya Kumar Mummadi
Thi-Phuong-Nhung Ngo
Thi Hoai Phuong Nguyen
Laura Beggel
Thomas Brox

Deep neural networks (DNNs) have been shown to over-fit a dataset when being trained with noisy labels for a long enough time. To overcome this problem, we present a simple and effective method self-ensemble label filtering (SELF) to progressively filter out the wrong labels during training. Our method improves the task performance by gradually allowing supervision only from the potentially non-noisy (clean) labels and stops learning on the filtered noisy labels. For the filtering, we form running averages of predictions over the entire training dataset using the network output at different training epochs. We show that these ensemble estimates yield more accurate identification of inconsistent predictions throughout training than the single estimates of the network at the most recent training epoch. While filtered samples are removed entirely from the supervised training loss, we dynamically leverage them via semi-supervised learning in the unsupervised loss. We demonstrate the positive effect of such an approach on various image classification tasks under both symmetric and asymmetric label noise and at different noise ratios. It substantially outperforms all previous works on noise-aware learning across different datasets and can be applied to a broad set of network architectures.

ICLR Conference 2020 Conference Paper

Understanding and Robustifying Differentiable Architecture Search

Arber Zela
Thomas Elsken
Tonmoy Saikia
Yassine Marrakchi
Thomas Brox
Frank Hutter

Differentiable Architecture Search (DARTS) has attracted a lot of attention due to its simplicity and small search costs achieved by a continuous relaxation and an approximation of the resulting bi-level optimization problem. However, DARTS does not work robustly for new problems: we identify a wide range of search spaces for which DARTS yields degenerate architectures with very poor test performance. We study this failure mode and show that, while DARTS successfully minimizes validation loss, the found solutions generalize poorly when they coincide with high validation loss curvature in the architecture space. We show that by adding one of various types of regularization we can robustify DARTS to find solutions with less curvature and better generalization properties. Based on these observations, we propose several simple variations of DARTS that perform substantially more robustly in practice. Our observations are robust across five search spaces on three image classification tasks and also hold for the very different domains of disparity estimation (a dense regression task) and language modelling.

ICML Conference 2019 Conference Paper

Anomaly Detection With Multiple-Hypotheses Predictions

Duc Tam Nguyen
Zhongyu Lou
Michael Klar
Thomas Brox

In one-class-learning tasks, only the normal case (foreground) can be modeled with data, whereas the variation of all possible anomalies is too erratic to be described by samples. Thus, due to the lack of representative data, the wide-spread discriminative approaches cannot cover such learning tasks, and rather generative models, which attempt to learn the input density of the foreground, are used. However, generative models suffer from a large input dimensionality (as in images) and are typically inefficient learners. We propose to learn the data distribution of the foreground more efficiently with a multi-hypotheses autoencoder. Moreover, the model is criticized by a discriminator, which prevents artificial data modes not supported by data, and which enforces diversity across hypotheses. Our multiple-hypotheses-based anomaly detection framework allows the reliable identification of out-of-distribution samples. For anomaly detection on CIFAR-10, it yields up to 3. 9% points improvement over previously reported results. On a real anomaly detection task, the approach reduces the error of the baseline models from 6. 8% to 1. 5%.

IROS Conference 2019 Conference Paper

Automated Boxwood Topiary Trimming with a Robotic Arm and Integrated Stereo Vision *

Dejan Kaljaca
Nikolaus Mayer
Bastiaan A. Vroegindeweij
Angelo Mencarelli
Eldert J. van Henten
Thomas Brox

This paper presents an integrated hardware-software solution to perform fully automated robotic bush trimming to user-specified shapes. In contrast to specialized solutions that can trim only bushes of a certain shape, the approach ensures flexibility via a vision-based shape fitting module that allows fitting an arbitrary mesh into a bush at hand. A trimming planning method considers the available degrees of freedom of the robot arm to achieve effective cutting motions. The performance of the mesh fitting module is assessed in multiple experiments involving both artificial and real plants with a variety of shapes. The trimming accuracy of the overall approach is quantitatively evaluated by inspecting the bush pointcloud before and after robotic trimming, and measuring the change in the deviation from the originally computed target mesh.

NeurIPS Conference 2019 Conference Paper

DeepUSPS: Deep Robust Unsupervised Saliency Prediction via Self-supervision

Tam Nguyen
Maximilian Dax
Chaithanya Kumar Mummadi
Nhung Ngo
Thi Hoai Phuong Nguyen
Zhongyu Lou
Thomas Brox

Deep neural network (DNN) based salient object detection in images based on high-quality labels is expensive. Alternative unsupervised approaches rely on careful selection of multiple handcrafted saliency methods to generate noisy pseudo-ground-truth labels. In this work, we propose a two-stage mechanism for robust unsupervised object saliency prediction, where the first stage involves refinement of the noisy pseudo labels generated from different handcrafted methods. Each handcrafted method is substituted by a deep network that learns to generate the pseudo labels. These labels are refined incrementally in multiple iterations via our proposed self-supervision technique. In the second stage, the refined labels produced from multiple networks representing multiple saliency methods are used to train the actual saliency detection network. We show that this self-learning procedure outperforms all the existing unsupervised methods over different datasets. Results are even comparable to those of fully-supervised state-of-the-art approaches.

IROS Conference 2019 Conference Paper

Self-supervised 3D Shape and Viewpoint Estimation from Single Images for Robotics

Oier Mees
Maxim Tatarchenko
Thomas Brox
Wolfram Burgard

We present a convolutional neural network for joint 3D shape prediction and viewpoint estimation from a single input image. During training, our network gets the learning signal from a silhouette of an object in the input image-a form of self-supervision. It does not require ground truth data for 3D shapes and the viewpoints. Because it relies on such a weak form of supervision, our approach can easily be applied to real-world data. We demonstrate that our method produces reasonable qualitative and quantitative results on natural images for both shape estimation and viewpoint prediction. Unlike previous approaches, our method does not require multiple views of the same object instance in the dataset, which significantly expands the applicability in practical robotics scenarios. We showcase it by using the hallucinated shapes to improve the performance on the task of grasping real-world objects both in simulation and with a PR2 robot.

ICRA Conference 2018 Conference Paper

3D Human Pose Estimation in RGBD Images for Robotic Task Learning

Christian Zimmermann 0001
Tim Welschehold
Christian Dornhege
Wolfram Burgard
Thomas Brox

We propose an approach to estimate 3D human pose in real world units from a single RGBD image and show that it exceeds performance of monocular 3D pose estimation approaches from color as well as pose estimation exclusively from depth. Our approach builds on robust human keypoint detectors for color images and incorporates depth for lifting into 3D. We combine the system with our learning from demonstration framework to instruct a service robot without the need of markers. Experiments in real world settings demonstrate that our approach enables a PR2 robot to imitate manipulation actions observed from a human teacher.

ICRA Conference 2018 Conference Paper

DPDB-Net: Exploiting Dense Connections for Convolutional Encoders

Gabriel L. Oliveira
Wolfram Burgard
Thomas Brox

Densely connected networks for classification enable feature exploration and result in state-of-the-art performance on multiple classification tasks. The alternative to dense networks is the residual network which enables feature re-usage. In this work, we combine these orthogonal concepts for encoder-decoder architectures, which we call Dual-Path Dense-Block Network (DPDB-Net). We introduce a dense block which incorporates feature re-usage and new feature exploration in the encoder. Moreover, we discuss that feature re-usage by the residual network architecture leads to a feature map explosion in the decoder and, thus, is not advantageous in this part of the network. We evaluated our proposed architecture in multiple segmentation tasks and report state-of-the-art performance on the Freiburg Forest dataset and competitive results on the Cam Vid dataset.

ICRA Conference 2017 Conference Paper

Semantics-aware visual localization under challenging perceptual conditions

Tayyab Naseer
Gabriel L. Oliveira
Thomas Brox
Wolfram Burgard

Visual place recognition under difficult perceptual conditions remains a challenging problem due to changing weather conditions, illumination and seasons. Long-term visual navigation approaches for robot localization should be robust to these dynamics of the environment. Existing methods typically leverage feature descriptions of whole images or image regions from Deep Convolutional Neural Networks. Some approaches also exploit sequential information to alleviate the problem of spatially inconsistent and non-perfect image matches. In this paper, we propose a novel approach for learning a discriminative holistic image representation which exploits the image content to create a dense and salient scene description. These salient descriptions are learnt over a variety of datasets under large perceptual changes. Such an approach enables us to precisely segment the regions of an image which are geometrically stable over large time lags. We combine features from these salient regions and an off-the-shelf holistic representation to form a more robust scene descriptor. We also introduce a semantically labeled dataset which captures extreme perceptual and structural scene dynamics over the course of 3 years. We evaluated our approach with extensive experiments on data collected over several kilometers in Freiburg and show that our learnt image representation outperforms off-the-shelf features from the deep networks and hand-crafted features.

ICRA Conference 2016 Conference Paper

Deep learning for human part discovery in images

Gabriel L. Oliveira
Abhinav Valada
Claas Bollen
Wolfram Burgard
Thomas Brox

This paper addresses the problem of human body part segmentation in conventional RGB images, which has several applications in robotics, such as learning from demonstration and human-robot handovers. The proposed solution is based on Convolutional Neural Networks (CNNs). We present a network architecture that assigns each pixel to one of a predefined set of human body part classes, such as head, torso, arms, legs. After initializing weights with a very deep convolutional network for image classification, the network can be trained end-to-end and yields precise class predictions at the original input resolution. Our architecture particularly improves on over-fitting issues in the up-convolutional part of the network. Relying only on RGB rather than RGB-D images also allows us to apply the approach outdoors. The network achieves state-of-the-art performance on the PASCAL Parts dataset. Moreover, we introduce two new part segmentation datasets, the Freiburg sitting people dataset and the Freiburg people in disaster dataset. We also present results obtained with a ground robot and an unmanned aerial vehicle.

IROS Conference 2016 Conference Paper

Efficient deep models for monocular road segmentation

Gabriel L. Oliveira
Wolfram Burgard
Thomas Brox

This paper addresses the problem of road scene segmentation in conventional RGB images by exploiting recent advances in semantic segmentation via convolutional neural networks (CNNs). Segmentation networks are very large and do not currently run at interactive frame rates. To make this technique applicable to robotics we propose several architecture refinements that provide the best trade-off between segmentation quality and runtime. This is achieved by a new mapping between classes and filters at the expansion side of the network. The network is trained end-to-end and yields precise road/lane predictions at the original input resolution in roughly 50ms. Compared to the state of the art, the network achieves top accuracies on the KITTI dataset for road and lane segmentation while providing a 20× speed-up. We demonstrate that the improved efficiency is not due to the road segmentation task. Also on segmentation datasets with larger scene complexity, the accuracy does not suffer from the large speed-up.

NeurIPS Conference 2016 Conference Paper

Generating Images with Perceptual Similarity Metrics based on Deep Networks

Alexey Dosovitskiy
Thomas Brox

We propose a class of loss functions, which we call deep perceptual similarity metrics (DeePSiM), allowing to generate sharp high resolution images from compressed abstract representations. Instead of computing distances in the image space, we compute distances between image features extracted by deep neural networks. This metric reflects perceptual similarity of images much better and, thus, leads to better results. We demonstrate two examples of use cases of the proposed loss: (1) networks that invert the AlexNet convolutional network; (2) a modified version of a variational autoencoder that generates realistic high-resolution random images.

NeurIPS Conference 2016 Conference Paper

Protein contact prediction from amino acid co-evolution using convolutional networks for graph-valued images

Vladimir Golkov
Marcin Skwark
Antonij Golkov
Alexey Dosovitskiy
Thomas Brox
Jens Meiler
Daniel Cremers

Proteins are the "building blocks of life", the most abundant organic molecules, and the central focus of most areas of biomedicine. Protein structure is strongly related to protein function, thus structure prediction is a crucial task on the way to solve many biological questions. A contact map is a compact representation of the three-dimensional structure of a protein via the pairwise contacts between the amino acid constituting the protein. We use a convolutional network to calculate protein contact maps from inferred statistical coupling between positions in the protein sequence. The input to the network has an image-like structure amenable to convolutions, but every "pixel" instead of color channels contains a bipartite undirected edge-weighted graph. We propose several methods for treating such "graph-valued images" in a convolutional network. The proposed method outperforms state-of-the-art methods by a large margin. It also allows for a great flexibility with regard to the input data, which makes it useful for studying a wide range of problems.

NeurIPS Conference 2016 Conference Paper

Synthesizing the preferred inputs for neurons in neural networks via deep generator networks

Anh Nguyen
Alexey Dosovitskiy
Jason Yosinski
Thomas Brox
Jeff Clune

Deep neural networks (DNNs) have demonstrated state-of-the-art results on many pattern recognition tasks, especially vision classification problems. Understanding the inner workings of such computational brains is both fascinating basic science that is interesting in its own right---similar to why we study the human brain---and will enable researchers to further improve DNNs. One path to understanding how a neural network functions internally is to study what each of its neurons has learned to detect. One such method is called activation maximization, which synthesizes an input (e. g. an image) that highly activates a neuron. Here we dramatically improve the qualitative state of the art of activation maximization by harnessing a powerful, learned prior: a deep generator network. The algorithm (1) generates qualitatively state-of-the-art synthetic images that look almost real, (2) reveals the features learned by each neuron in an interpretable way, (3) generalizes well to new datasets and somewhat well to different network architectures without requiring the prior to be relearned, and (4) can be considered as a high-quality generative method (in this case, by generating novel, creative, interesting, recognizable images).

NeurIPS Conference 2014 Conference Paper

Discriminative Unsupervised Feature Learning with Convolutional Neural Networks

Alexey Dosovitskiy
Jost Tobias Springenberg
Martin Riedmiller
Thomas Brox

Current methods for training convolutional neural networks depend on large amounts of labeled samples for supervised training. In this paper we present an approach for training a convolutional neural network using only unlabeled data. We train the network to discriminate between a set of surrogate classes. Each surrogate class is formed by applying a variety of transformations to a randomly sampled 'seed' image patch. We find that this simple feature learning algorithm is surprisingly successful when applied to visual object recognition. The feature representation learned by our algorithm achieves classification results matching or outperforming the current state-of-the-art for unsupervised learning on several popular datasets (STL-10, CIFAR-10, Caltech-101).

ICRA Conference 2014 Conference Paper

Reconstruction of rigid body models from motion distorted laser range data using optical flow

Eddy Ilg
Rainer Kümmerle
Wolfram Burgard
Thomas Brox

The setup of tilting a 2D laser range finder up and down is a widespread strategy to acquire 3D point clouds. This setup requires that the scene is static while the robot takes a 3D scan. If an object moves through the scene during the measurement process and one does not take into account these movements, the resulting model will get distorted. This paper presents an approach to reconstruct the 3D model of a moving rigid object from the inconsistent set of 2D measurements by the help of a camera. Our approach utilizes optical flow in the camera images to estimate the motion in the image plane and point-line constraints to compensate the missing information about the motion in depth. We combine multiple sweeps and/or views into to a single consistent model using a point-to-plane ICP approach and optimize single sweeps by smoothing the resulting trajectory. Experiments obtained in real outdoor scenarios with moving cars demonstrate that our approach yields accurate models.