Samuel Schulter

Shiyu Zhao, Zhixing Zhang, Samuel Schulter et al. · deepmind

Building robust and generic object detection frameworks requires scaling to larger label spaces and bigger training datasets. However, it is prohibitively costly to acquire annotations for thousands of categories at a large scale. We propose a novel method that leverages the rich semantics available in recent vision and language models to localize and classify objects in unlabeled images, effectively generating pseudo labels for object detection. Starting with a generic and class-agnostic region proposal mechanism, we use vision and language models to categorize each region of an image into any object category that is required for downstream tasks. We demonstrate the value of the generated pseudo labels in two specific tasks, open-vocabulary detection, where a model needs to generalize to unseen object categories, and semi-supervised object detection, where additional unlabeled images can be used to improve the model. Our empirical evaluation shows the effectiveness of the pseudo labels in both tasks, where we outperform competitive baselines and achieve a novel state-of-the-art for open-vocabulary object detection. Our code is available at https://github.com/xiaofeng94/VL-PLM.

Shiyu Zhao, Samuel Schulter, Long Zhao et al. · deepmind

Recent studies have shown promising performance in open-vocabulary object detection (OVD) by utilizing pseudo labels (PLs) from pretrained vision and language models (VLMs). However, teacher-student self-training, a powerful and widely used paradigm to leverage PLs, is rarely explored for OVD. This work identifies two challenges of using self-training in OVD: noisy PLs from VLMs and frequent distribution changes of PLs. To address these challenges, we propose SAS-Det that tames self-training for OVD from two key perspectives. First, we present a split-and-fusion (SAF) head that splits a standard detection into an open-branch and a closed-branch. This design can reduce noisy supervision from pseudo boxes. Moreover, the two branches learn complementary knowledge from different training data, significantly enhancing performance when fused together. Second, in our view, unlike in closed-set tasks, the PL distributions in OVD are solely determined by the teacher model. We introduce a periodic update strategy to decrease the number of updates to the teacher, thereby decreasing the frequency of changes in PL distributions, which stabilizes the training process. Extensive experiments demonstrate SAS-Det is both efficient and effective. SAS-Det outperforms recent models of the same scale by a clear margin and achieves 37.4 AP50 and 29.1 APr on novel categories of the COCO and LVIS benchmarks, respectively. Code is available at \url{https://github.com/xiaofeng94/SAS-Det}.

Zaid Khan, Vijay Kumar BG, Samuel Schulter et al.

Finetuning a large vision language model (VLM) on a target dataset after large scale pretraining is a dominant paradigm in visual question answering (VQA). Datasets for specialized tasks such as knowledge-based VQA or VQA in non natural-image domains are orders of magnitude smaller than those for general-purpose VQA. While collecting additional labels for specialized tasks or domains can be challenging, unlabeled images are often available. We introduce SelTDA (Self-Taught Data Augmentation), a strategy for finetuning large VLMs on small-scale VQA datasets. SelTDA uses the VLM and target dataset to build a teacher model that can generate question-answer pseudolabels directly conditioned on an image alone, allowing us to pseudolabel unlabeled images. SelTDA then finetunes the initial VLM on the original dataset augmented with freshly pseudolabeled images. We describe a series of experiments showing that our self-taught data augmentation increases robustness to adversarially searched questions, counterfactual examples and rephrasings, improves domain generalization, and results in greater retention of numerical reasoning skills. The proposed strategy requires no additional annotations or architectural modifications, and is compatible with any modern encoder-decoder multimodal transformer. Code available at https://github.com/codezakh/SelTDA.

Amazon AGI, Aaron Langford, Aayush Shah et al. · amazon-science

We present Amazon Nova, a new generation of state-of-the-art foundation models that deliver frontier intelligence and industry-leading price performance. Amazon Nova Pro is a highly-capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks. Amazon Nova Lite is a low-cost multimodal model that is lightning fast for processing images, video, documents and text. Amazon Nova Micro is a text-only model that delivers our lowest-latency responses at very low cost. Amazon Nova Canvas is an image generation model that creates professional grade images with rich customization controls. Amazon Nova Reel is a video generation model offering high-quality outputs, customization, and motion control. Our models were built responsibly and with a commitment to customer trust, security, and reliability. We report benchmarking results for core capabilities, agentic performance, long context, functional adaptation, runtime performance, and human evaluation.

Samuel Schulter, Vijay Kumar B G, Yumin Suh et al.

Language-based object detection is a promising direction towards building a natural interface to describe objects in images that goes far beyond plain category names. While recent methods show great progress in that direction, proper evaluation is lacking. With OmniLabel, we propose a novel task definition, dataset, and evaluation metric. The task subsumes standard- and open-vocabulary detection as well as referring expressions. With more than 28K unique object descriptions on over 25K images, OmniLabel provides a challenging benchmark with diverse and complex object descriptions in a naturally open-vocabulary setting. Moreover, a key differentiation to existing benchmarks is that our object descriptions can refer to one, multiple or even no object, hence, providing negative examples in free-form text. The proposed evaluation handles the large label space and judges performance via a modified average precision metric, which we validate by evaluating strong language-based baselines. OmniLabel indeed provides a challenging test bed for future research on language-based detection.

Zaid Khan, Vijay Kumar BG, Xiang Yu et al.

Self-supervised vision-language pretraining from pure images and text with a contrastive loss is effective, but ignores fine-grained alignment due to a dual-stream architecture that aligns image and text representations only on a global level. Earlier, supervised, non-contrastive methods were capable of finer-grained alignment, but required dense annotations that were not scalable. We propose a single stream architecture that aligns images and language at multiple levels: global, fine-grained patch-token, and conceptual/semantic, using two novel tasks: symmetric cross-modality reconstruction (XMM) and a pseudo-labeled key word prediction (PSL). In XMM, we mask input tokens from one modality and use cross-modal information to reconstruct the masked token, thus improving fine-grained alignment between the two modalities. In PSL, we use attention to select keywords in a caption, use a momentum encoder to recommend other important keywords that are missing from the caption but represented in the image, and then train the visual encoder to predict the presence of those keywords, helping it learn semantic concepts that are essential for grounding a textual token to an image region. We demonstrate competitive performance and improved data efficiency on image-text retrieval, grounding, visual question answering/reasoning against larger models and models trained on more data. Code and models available at zaidkhan.me/SIMLA.

Dripta S. Raychaudhuri, Yumin Suh, Samuel Schulter et al.

Multi-task learning commonly encounters competition for resources among tasks, specifically when model capacity is limited. This challenge motivates models which allow control over the relative importance of tasks and total compute cost during inference time. In this work, we propose such a controllable multi-task network that dynamically adjusts its architecture and weights to match the desired task preference as well as the resource constraints. In contrast to the existing dynamic multi-task approaches that adjust only the weights within a fixed architecture, our approach affords the flexibility to dynamically control the total computational cost and match the user-preferred task importance better. We propose a disentangled training of two hypernetworks, by exploiting task affinity and a novel branching regularized loss, to take input preferences and accordingly predict tree-structured models with adapted weights. Experiments on three multi-task benchmarks, namely PASCAL-Context, NYU-v2, and CIFAR-100, show the efficacy of our approach. Project page is available at https://www.nec-labs.com/~mas/DYMU.

Christian Simon, Masoud Faraki, Yi-Hsuan Tsai et al.

Humans have the ability to accumulate knowledge of new tasks in varying conditions, but deep neural networks often suffer from catastrophic forgetting of previously learned knowledge after learning a new task. Many recent methods focus on preventing catastrophic forgetting under the assumption of train and test data following similar distributions. In this work, we consider a more realistic scenario of continual learning under domain shifts where the model must generalize its inference to an unseen domain. To this end, we encourage learning semantically meaningful features by equipping the classifier with class similarity metrics as learning parameters which are obtained through Mahalanobis similarity computations. Learning of the backbone representation along with these extra parameters is done seamlessly in an end-to-end manner. In addition, we propose an approach based on the exponential moving average of the parameters for better knowledge distillation. We demonstrate that, to a great extent, existing continual learning algorithms fail to handle the forgetting issue under multiple distributions, while our proposed approach learns new tasks under domain shift with accuracy boosts up to 10% on challenging datasets such as DomainNet and OfficeHome.

Inkyu Shin, Yi-Hsuan Tsai, Bingbing Zhuang et al.

Test-time adaptation approaches have recently emerged as a practical solution for handling domain shift without access to the source domain data. In this paper, we propose and explore a new multi-modal extension of test-time adaptation for 3D semantic segmentation. We find that directly applying existing methods usually results in performance instability at test time because multi-modal input is not considered jointly. To design a framework that can take full advantage of multi-modality, where each modality provides regularized self-supervisory signals to other modalities, we propose two complementary modules within and across the modalities. First, Intra-modal Pseudolabel Generation (Intra-PG) is introduced to obtain reliable pseudo labels within each modality by aggregating information from two models that are both pre-trained on source data but updated with target data at different paces. Second, Inter-modal Pseudo-label Refinement (Inter-PR) adaptively selects more reliable pseudo labels from different modalities based on a proposed consistency scheme. Experiments demonstrate that our regularized pseudo labels produce stable self-learning signals in numerous multi-modal test-time adaptation scenarios for 3D semantic segmentation. Visit our project website at https://www.nec-labs.com/~mas/MM-TTA.

Abhishek Aich, Samuel Schulter, Amit K. Roy-Chowdhury et al.

We aim to train a multi-task model such that users can adjust the desired compute budget and relative importance of task performances after deployment, without retraining. This enables optimizing performance for dynamically varying user needs, without heavy computational overhead to train and save models for various scenarios. To this end, we propose a multi-task model consisting of a shared encoder and task-specific decoders where both encoder and decoder channel widths are slimmable. Our key idea is to control the task importance by varying the capacities of task-specific decoders, while controlling the total computational cost by jointly adjusting the encoder capacity. This improves overall accuracy by allowing a stronger encoder for a given budget, increases control over computational cost, and delivers high-quality slimmed sub-architectures based on user's constraints. Our training strategy involves a novel 'Configuration-Invariant Knowledge Distillation' loss that enforces backbone representations to be invariant under different runtime width configurations to enhance accuracy. Further, we present a simple but effective search algorithm that translates user constraints to runtime width configurations of both the shared encoder and task decoders, for sampling the sub-architectures. The key rule for the search algorithm is to provide a larger computational budget to the higher preferred task decoder, while searching a shared encoder configuration that enhances the overall MTL performance. Various experiments on three multi-task benchmarks (PASCALContext, NYUDv2, and CIFAR100-MTL) with diverse backbone architectures demonstrate the advantage of our approach. For example, our method shows a higher controllability by ~33.5% in the NYUD-v2 dataset over prior methods, while incurring much less compute cost.

Zaid Khan, Vijay Kumar BG, Samuel Schulter et al.

Visual question answering (VQA) has traditionally been treated as a single-step task where each question receives the same amount of effort, unlike natural human question-answering strategies. We explore a question decomposition strategy for VQA to overcome this limitation. We probe the ability of recently developed large vision-language models to use human-written decompositions and produce their own decompositions of visual questions, finding they are capable of learning both tasks from demonstrations alone. However, we show that naive application of model-written decompositions can hurt performance. We introduce a model-driven selective decomposition approach for second-guessing predictions and correcting errors, and validate its effectiveness on eight VQA tasks across three domains, showing consistent improvements in accuracy, including improvements of >20% on medical VQA datasets and boosting the zero-shot performance of BLIP-2 above chance on a VQA reformulation of the challenging Winoground task. Project Site: https://zaidkhan.me/decomposition-0shot-vqa/

Qilong Zhangli, Di Liu, Abhishek Aich et al.

Leveraging multiple training datasets to scale up image segmentation models is beneficial for increasing robustness and semantic understanding. Individual datasets have well-defined ground truth with non-overlapping mask layouts and mutually exclusive semantics. However, merging them for multi-dataset training disrupts this harmony and leads to semantic inconsistencies; for example, the class "person" in one dataset and class "face" in another will require multilabel handling for certain pixels. Existing methods struggle with this setting, particularly when evaluated on label spaces mixed from the individual training sets. To overcome these issues, we introduce a simple yet effective multi-dataset training approach by integrating language-based embeddings of class names and label space-specific query embeddings. Our method maintains high performance regardless of the underlying inconsistencies between training datasets. Notably, on four benchmark datasets with label space inconsistencies during inference, we outperform previous methods by 1.6% mIoU for semantic segmentation, 9.1% PQ for panoptic segmentation, 12.1% AP for instance segmentation, and 3.0% in the newly proposed PIQ metric.

Sirui Xu, Samuel Schulter, Morteza Ziyadi et al.

Humans rarely plan whole-body interactions with objects at the level of explicit whole-body movements. High-level intentions, such as affordance, define the goal, while coordinated balance, contact, and manipulation can emerge naturally from underlying physical and motor priors. Scaling such priors is key to enabling humanoids to compose and generalize loco-manipulation skills across diverse contexts while maintaining physically coherent whole-body coordination. To this end, we introduce InterPrior, a scalable framework that learns a unified generative controller through large-scale imitation pretraining and post-training by reinforcement learning. InterPrior first distills a full-reference imitation expert into a versatile, goal-conditioned variational policy that reconstructs motion from multimodal observations and high-level intent. While the distilled policy reconstructs training behaviors, it does not generalize reliably due to the vast configuration space of large-scale human-object interactions. To address this, we apply data augmentation with physical perturbations, and then perform reinforcement learning finetuning to improve competence on unseen goals and initializations. Together, these steps consolidate the reconstructed latent skills into a valid manifold, yielding a motion prior that generalizes beyond the training data, e.g., it can incorporate new behaviors such as interactions with unseen objects. We further demonstrate its effectiveness for user-interactive control and its potential for real robot deployment.

Shiyu Zhao, Long Zhao, Vijay Kumar B. G et al. · deepmind

The recent progress in language-based open-vocabulary object detection can be largely attributed to finding better ways of leveraging large-scale data with free-form text annotations. Training such models with a discriminative objective function has proven successful, but requires good positive and negative samples. However, the free-form nature and the open vocabulary of object descriptions make the space of negatives extremely large. Prior works randomly sample negatives or use rule-based techniques to build them. In contrast, we propose to leverage the vast knowledge built into modern generative models to automatically build negatives that are more relevant to the original data. Specifically, we use large-language-models to generate negative text descriptions, and text-to-image diffusion models to also generate corresponding negative images. Our experimental analysis confirms the relevance of the generated negative data, and its use in language-based detectors improves performance on two complex benchmarks. Code is available at \url{https://github.com/xiaofeng94/Gen-Enhanced-Negs}.

Abhishek Aich, Yumin Suh, Samuel Schulter et al.

A powerful architecture for universal segmentation relies on transformers that encode multi-scale image features and decode object queries into mask predictions. With efficiency being a high priority for scaling such models, we observed that the state-of-the-art method Mask2Former uses 50% of its compute only on the transformer encoder. This is due to the retention of a full-length token-level representation of all backbone feature scales at each encoder layer. With this observation, we propose a strategy termed PROgressive Token Length SCALing for Efficient transformer encoders (PRO-SCALE) that can be plugged-in to the Mask2Former segmentation architecture to significantly reduce the computational cost. The underlying principle of PRO-SCALE is: progressively scale the length of the tokens with the layers of the encoder. This allows PRO-SCALE to reduce computations by a large margin with minimal sacrifice in performance (~52% encoder and ~27% overall GFLOPs reduction with no drop in performance on COCO dataset). Experiments conducted on public benchmarks demonstrates PRO-SCALE's flexibility in architectural configurations, and exhibits potential for extension beyond the settings of segmentation tasks to encompass object detection. Code here: https://github.com/abhishekaich27/proscale-pytorch

Mingfu Liang, Jong-Chyi Su, Samuel Schulter et al.

Autonomous vehicle (AV) systems rely on robust perception models as a cornerstone of safety assurance. However, objects encountered on the road exhibit a long-tailed distribution, with rare or unseen categories posing challenges to a deployed perception model. This necessitates an expensive process of continuously curating and annotating data with significant human effort. We propose to leverage recent advances in vision-language and large language models to design an Automatic Data Engine (AIDE) that automatically identifies issues, efficiently curates data, improves the model through auto-labeling, and verifies the model through generation of diverse scenarios. This process operates iteratively, allowing for continuous self-improvement of the model. We further establish a benchmark for open-world detection on AV datasets to comprehensively evaluate various learning paradigms, demonstrating our method's superior performance at a reduced cost.

Zaid Khan, Vijay Kumar BG, Samuel Schulter et al.

Visual program synthesis is a promising approach to exploit the reasoning abilities of large language models for compositional computer vision tasks. Previous work has used few-shot prompting with frozen LLMs to synthesize visual programs. Training an LLM to write better visual programs is an attractive prospect, but it is unclear how to accomplish this. No dataset of visual programs for training exists, and acquisition of a visual program dataset cannot be easily crowdsourced due to the need for expert annotators. To get around the lack of direct supervision, we explore improving the program synthesis abilities of an LLM using feedback from interactive experience. We propose a method where we exploit existing annotations for a vision-language task to improvise a coarse reward signal for that task, treat the LLM as a policy, and apply reinforced self-training to improve the visual program synthesis ability of the LLM for that task. We describe a series of experiments on object detection, compositional visual question answering, and image-text retrieval, and show that in each case, the self-trained LLM outperforms or performs on par with few-shot frozen LLMs that are an order of magnitude larger. Website: https://zaidkhan.me/ViReP

David Schinagl, Christian Fruhwirth-Reisinger, Alexander Prutsch et al.

Interpreting the decisions of complex computer vision models is crucial to establish trust and accountability, especially in safety-critical domains. An established approach to interpretability is generating visual attribution maps that highlight regions of the input most relevant to the model's prediction. However, existing methods face a three-way trade-off. Propagation-based approaches are efficient, but they can be biased and architecture-specific. Meanwhile, perturbation-based methods are causally grounded, yet they are expensive and for vision transformers often yield coarse, patch-level explanations. Learning-based explainers are fast but usually optimize surrogate objectives or distill from heuristic teachers. We propose a learning scheme that instead optimizes deletion and insertion metrics directly. Since these metrics depend on non-differentiable sorting and ranking, we frame them as permutation learning and replace the hard sorting with a differentiable relaxation using Gumbel-Sinkhorn. This enables end-to-end training through attribution-guided perturbations of the target model. During inference, our method produces dense, pixel-level attributions in a single forward pass with optional, few-step gradient refinement. Our experiments demonstrate consistent quantitative improvements and sharper, boundary-aligned explanations, particularly for transformer-based vision models.

Dušan Malić, Christian Fruhwirth-Reisinger, Samuel Schulter et al.

LiDAR-based 3D detectors need large datasets for training, yet they struggle to generalize to novel domains. Domain Generalization (DG) aims to mitigate this by training detectors that are invariant to such domain shifts. Current DG approaches exclusively rely on global geometric features (point cloud Cartesian coordinates) as input features. Over-reliance on these global geometric features can, however, cause 3D detectors to prioritize object location and absolute position, resulting in poor cross-domain performance. To mitigate this, we propose to exploit explicit local point cloud structure for DG, in particular by encoding point cloud neighborhoods with Gaussian blobs, GBlobs. Our proposed formulation is highly efficient and requires no additional parameters. Without any bells and whistles, simply by integrating GBlobs in existing detectors, we beat the current state-of-the-art in challenging single-source DG benchmarks by over 21 mAP (Waymo->KITTI), 13 mAP (KITTI->Waymo), and 12 mAP (nuScenes->KITTI), without sacrificing in-domain performance. Additionally, GBlobs demonstrate exceptional performance in multi-source DG, surpassing the current state-of-the-art by 17, 12, and 5 mAP on Waymo, KITTI, and ONCE, respectively.

Dušan Malić, Christian Fruhwirth-Reisinger, Samuel Schulter et al.

While surface normals are widely used to analyse 3D scene geometry, surface normal estimation from LiDAR point clouds remains severely underexplored. This is caused by the lack of large-scale annotated datasets on the one hand, and lack of methods that can robustly handle the sparse and often noisy LiDAR data in a reasonable time on the other hand. We address these limitations using a traffic simulation engine and present LiSu, the first large-scale, synthetic LiDAR point cloud dataset with ground truth surface normal annotations, eliminating the need for tedious manual labeling. Additionally, we propose a novel method that exploits the spatiotemporal characteristics of autonomous driving data to enhance surface normal estimation accuracy. By incorporating two regularization terms, we enforce spatial consistency among neighboring points and temporal smoothness across consecutive LiDAR frames. These regularizers are particularly effective in self-training settings, where they mitigate the impact of noisy pseudo-labels, enabling robust real-world deployment. We demonstrate the effectiveness of our method on LiSu, achieving state-of-the-art performance in LiDAR surface normal estimation. Moreover, we showcase its full potential in addressing the challenging task of synthetic-to-real domain adaptation, leading to improved neural surface reconstruction on real-world data.

Dušan Malić, Christian Fruhwirth-Reisinger, Alexander Prutsch et al.

This technical report outlines the top-ranking solution for RoboSense 2025: Track 3, achieving state-of-the-art performance on 3D object detection under various sensor placements. Our submission utilizes GBlobs, a local point cloud feature descriptor specifically designed to enhance model generalization across diverse LiDAR configurations. Current LiDAR-based 3D detectors often suffer from a \enquote{geometric shortcut} when trained on conventional global features (\ie, absolute Cartesian coordinates). This introduces a position bias that causes models to primarily rely on absolute object position rather than distinguishing shape and appearance characteristics. Although effective for in-domain data, this shortcut severely limits generalization when encountering different point distributions, such as those resulting from varying sensor placements. By using GBlobs as network input features, we effectively circumvent this geometric shortcut, compelling the network to learn robust, object-centric representations. This approach significantly enhances the model's ability to generalize, resulting in the exceptional performance demonstrated in this challenge.

Zhixiang Min, Bingbing Zhuang, Samuel Schulter et al.

Monocular 3D object localization in driving scenes is a crucial task, but challenging due to its ill-posed nature. Estimating 3D coordinates for each pixel on the object surface holds great potential as it provides dense 2D-3D geometric constraints for the underlying PnP problem. However, high-quality ground truth supervision is not available in driving scenes due to sparsity and various artifacts of Lidar data, as well as the practical infeasibility of collecting per-instance CAD models. In this work, we present NeurOCS, a framework that uses instance masks and 3D boxes as input to learn 3D object shapes by means of differentiable rendering, which further serves as supervision for learning dense object coordinates. Our approach rests on insights in learning a category-level shape prior directly from real driving scenes, while properly handling single-view ambiguities. Furthermore, we study and make critical design choices to learn object coordinates more effectively from an object-centric view. Altogether, our framework leads to new state-of-the-art in monocular 3D localization that ranks 1st on the KITTI-Object benchmark among published monocular methods.

Weizhe Liu, David Ferstl, Samuel Schulter et al.

We introduce a novel approach to unsupervised and semi-supervised domain adaptation for semantic segmentation. Unlike many earlier methods that rely on adversarial learning for feature alignment, we leverage contrastive learning to bridge the domain gap by aligning the features of structurally similar label patches across domains. As a result, the networks are easier to train and deliver better performance. Our approach consistently outperforms state-of-the-art unsupervised and semi-supervised methods on two challenging domain adaptive segmentation tasks, particularly with a small number of target domain annotations. It can also be naturally extended to weakly-supervised domain adaptation, where only a minor drop in accuracy can save up to 75% of annotation cost.

Xiangyun Zhao, Samuel Schulter, Gaurav Sharma et al.

Given multiple datasets with different label spaces, the goal of this work is to train a single object detector predicting over the union of all the label spaces. The practical benefits of such an object detector are obvious and significant application-relevant categories can be picked and merged form arbitrary existing datasets. However, naive merging of datasets is not possible in this case, due to inconsistent object annotations. Consider an object category like faces that is annotated in one dataset, but is not annotated in another dataset, although the object itself appears in the latter images. Some categories, like face here, would thus be considered foreground in one dataset, but background in another. To address this challenge, we design a framework which works with such partial annotations, and we exploit a pseudo labeling approach that we adapt for our specific case. We propose loss functions that carefully integrate partial but correct annotations with complementary but noisy pseudo labels. Evaluation in the proposed novel setting requires full annotation on the test set. We collect the required annotations and define a new challenging experimental setup for this task based one existing public datasets. We show improved performances compared to competitive baselines and appropriate adaptations of existing work.

Sujoy Paul, Yi-Hsuan Tsai, Samuel Schulter et al.

Learning semantic segmentation models requires a huge amount of pixel-wise labeling. However, labeled data may only be available abundantly in a domain different from the desired target domain, which only has minimal or no annotations. In this work, we propose a novel framework for domain adaptation in semantic segmentation with image-level weak labels in the target domain. The weak labels may be obtained based on a model prediction for unsupervised domain adaptation (UDA), or from a human annotator in a new weakly-supervised domain adaptation (WDA) paradigm for semantic segmentation. Using weak labels is both practical and useful, since (i) collecting image-level target annotations is comparably cheap in WDA and incurs no cost in UDA, and (ii) it opens the opportunity for category-wise domain alignment. Our framework uses weak labels to enable the interplay between feature alignment and pseudo-labeling, improving both in the process of domain adaptation. Specifically, we develop a weak-label classification module to enforce the network to attend to certain categories, and then use such training signals to guide the proposed category-wise alignment method. In experiments, we show considerable improvements with respect to the existing state-of-the-arts in UDA and present a new benchmark in the WDA setting. Project page is at http://www.nec-labs.com/~mas/WeakSegDA.

Buyu Liu, Bingbing Zhuang, Samuel Schulter et al.

In this paper, we address the problem of inferring the layout of complex road scenes from video sequences. To this end, we formulate it as a top-view road attributes prediction problem and our goal is to predict these attributes for each frame both accurately and consistently. In contrast to prior work, we exploit the following three novel aspects: leveraging camera motions in videos, including context cuesand incorporating long-term video information. Specifically, we introduce a model that aims to enforce prediction consistency in videos. Our model consists of one LSTM and one Feature Transform Module (FTM). The former implicitly incorporates the consistency constraint with its hidden states, and the latter explicitly takes the camera motion into consideration when aggregating information along videos. Moreover, we propose to incorporate context information by introducing road participants, e.g. objects, into our model. When the entire video sequence is available, our model is also able to encode both local and global cues, e.g. information from both past and future frames. Experiments on two data sets show that: (1) Incorporating either globalor contextual cues improves the prediction accuracy and leveraging both gives the best performance. (2) Introducing the LSTM and FTM modules improves the prediction consistency in videos. (3) The proposed method outperforms the SOTA by a large margin.

Yi-Hsuan Tsai, Kihyuk Sohn, Samuel Schulter et al.

Predicting structured outputs such as semantic segmentation relies on expensive per-pixel annotations to learn supervised models like convolutional neural networks. However, models trained on one data domain may not generalize well to other domains without annotations for model finetuning. To avoid the labor-intensive process of annotation, we develop a domain adaptation method to adapt the source data to the unlabeled target domain. We propose to learn discriminative feature representations of patches in the source domain by discovering multiple modes of patch-wise output distribution through the construction of a clustered space. With such representations as guidance, we use an adversarial learning scheme to push the feature representations of target patches in the clustered space closer to the distributions of source patches. In addition, we show that our framework is complementary to existing domain adaptation techniques and achieves consistent improvements on semantic segmentation. Extensive ablations and results are demonstrated on numerous benchmark datasets with various settings, such as synthetic-to-real and cross-city scenarios.

Ziyan Wang, Buyu Liu, Samuel Schulter et al.

In this paper, we address the problem of inferring the layout of complex road scenes given a single camera as input. To achieve that, we first propose a novel parameterized model of road layouts in a top-view representation, which is not only intuitive for human visualization but also provides an interpretable interface for higher-level decision making. Moreover, the design of our top-view scene model allows for efficient sampling and thus generation of large-scale simulated data, which we leverage to train a deep neural network to infer our scene model's parameters. Specifically, our proposed training procedure uses supervised domain-adaptation techniques to incorporate both simulated as well as manually annotated data. Finally, we design a Conditional Random Field (CRF) that enforces coherent predictions for a single frame and encourages temporal smoothness among video frames. Experiments on two public data sets show that: (1) Our parametric top-view model is representative enough to describe complex road scenes; (2) The proposed method outperforms baselines trained on manually-annotated or simulated data only, thus getting the best of both; (3) Our CRF is able to generate temporally smoothed while semantically meaningful results.

Nataniel Ruiz, Samuel Schulter, Manmohan Chandraker

Simulation is a useful tool in situations where training data for machine learning models is costly to annotate or even hard to acquire. In this work, we propose a reinforcement learning-based method for automatically adjusting the parameters of any (non-differentiable) simulator, thereby controlling the distribution of synthesized data in order to maximize the accuracy of a model trained on that data. In contrast to prior art that hand-crafts these simulation parameters or adjusts only parts of the available parameters, our approach fully controls the simulator with the actual underlying goal of maximizing accuracy, rather than mimicking the real data distribution or randomly generating a large volume of data. We find that our approach (i) quickly converges to the optimal simulation parameters in controlled experiments and (ii) can indeed discover good sets of parameters for an image rendering simulator in actual computer vision applications.

Samuel Schulter, Menghua Zhai, Nathan Jacobs et al.

Given a single RGB image of a complex outdoor road scene in the perspective view, we address the novel problem of estimating an occlusion-reasoned semantic scene layout in the top-view. This challenging problem not only requires an accurate understanding of both the 3D geometry and the semantics of the visible scene, but also of occluded areas. We propose a convolutional neural network that learns to predict occluded portions of the scene layout by looking around foreground objects like cars or pedestrians. But instead of hallucinating RGB values, we show that directly predicting the semantics and depths in the occluded areas enables a better transformation into the top-view. We further show that this initial top-view representation can be significantly enhanced by learning priors and rules about typical road layouts from simulated or, if available, map data. Crucially, training our model does not require costly or subjective human annotations for occluded areas or the top-view, but rather uses readily available annotations for standard semantic segmentation. We extensively evaluate and analyze our approach on the KITTI and Cityscapes data sets.

Tuan-Hung Vu, Wongun Choi, Samuel Schulter et al.

This paper proposes a novel memory-based online video representation that is efficient, accurate and predictive. This is in contrast to prior works that often rely on computationally heavy 3D convolutions, ignore actual motion when aligning features over time, or operate in an off-line mode to utilize future frames. In particular, our memory (i) holds the feature representation, (ii) is spatially warped over time to compensate for observer and scene motions, (iii) can carry long-term information, and (iv) enables predicting feature representations in future frames. By exploring a variant that operates at multiple temporal scales, we efficiently learn across even longer time horizons. We apply our online framework to object detection in videos, obtaining a large 2.3 times speed-up and losing only 0.9% mAP on ImageNet-VID dataset, compared to prior works that even use future frames. Finally, we demonstrate the predictive property of our representation in two novel detection setups, where features are propagated over time to (i) significantly enhance a real-time detector by more than 10% mAP in a multi-threaded online setup and to (ii) anticipate objects in future frames.

Yi-Hsuan Tsai, Wei-Chih Hung, Samuel Schulter et al.

Convolutional neural network-based approaches for semantic segmentation rely on supervision with pixel-level ground truth, but may not generalize well to unseen image domains. As the labeling process is tedious and labor intensive, developing algorithms that can adapt source ground truth labels to the target domain is of great interest. In this paper, we propose an adversarial learning method for domain adaptation in the context of semantic segmentation. Considering semantic segmentations as structured outputs that contain spatial similarities between the source and target domains, we adopt adversarial learning in the output space. To further enhance the adapted model, we construct a multi-level adversarial network to effectively perform output space domain adaptation at different feature levels. Extensive experiments and ablation study are conducted under various domain adaptation settings, including synthetic-to-real and cross-city scenarios. We show that the proposed method performs favorably against the state-of-the-art methods in terms of accuracy and visual quality.

Samuel Schulter, Paul Vernaza, Wongun Choi et al.

Data association problems are an important component of many computer vision applications, with multi-object tracking being one of the most prominent examples. A typical approach to data association involves finding a graph matching or network flow that minimizes a sum of pairwise association costs, which are often either hand-crafted or learned as linear functions of fixed features. In this work, we demonstrate that it is possible to learn features for network-flow-based data association via backpropagation, by expressing the optimum of a smoothed network flow problem as a differentiable function of the pairwise association costs. We apply this approach to multi-object tracking with a network flow formulation. Our experiments demonstrate that we are able to successfully learn all cost functions for the association problem in an end-to-end fashion, which outperform hand-crafted costs in all settings. The integration and combination of various sources of inputs becomes easy and the cost functions can be learned entirely from data, alleviating tedious hand-designing of costs.

Georg Poier, Konstantinos Roditakis, Samuel Schulter et al.

Model-based approaches to 3D hand tracking have been shown to perform well in a wide range of scenarios. However, they require initialisation and cannot recover easily from tracking failures that occur due to fast hand motions. Data-driven approaches, on the other hand, can quickly deliver a solution, but the results often suffer from lower accuracy or missing anatomical validity compared to those obtained from model-based approaches. In this work we propose a hybrid approach for hand pose estimation from a single depth image. First, a learned regressor is employed to deliver multiple initial hypotheses for the 3D position of each hand joint. Subsequently, the kinematic parameters of a 3D hand model are found by deliberately exploiting the inherent uncertainty of the inferred joint proposals. This way, the method provides anatomically valid and accurate solutions without requiring manual initialisation or suffering from track losses. Quantitative results on several standard datasets demonstrate that the proposed method outperforms state-of-the-art representatives of the model-based, data-driven and hybrid paradigms.