Determining the relative pose of an object between two images is pivotal to the success of generalizable object pose estimation. Existing approaches typically approximate the continuous pose representation with a large number of discrete pose hypotheses, which incurs a computationally expensive process of scoring each hypothesis at test time. By contrast, we present a Deep Voxel Matching Network (DVMNet) that eliminates the need for pose hypotheses and computes the relative object pose in a single pass. To this end, we map the two input RGB images, reference and query, to their respective voxelized 3D representations. We then pass the resulting voxels through a pose estimation module, where the voxels are aligned and the pose is computed in an end-to-end fashion by solving a least-squares problem. To enhance robustness, we introduce a weighted closest voxel algorithm capable of mitigating the impact of noisy voxels. We conduct extensive experiments on the CO3D, LINEMOD, and Objaverse datasets, demonstrating that our method delivers more accurate relative pose estimates for novel objects at a lower computational cost compared to state-of-the-art methods. Our code is released at: https://github.com/sailor-z/DVMNet/.

When enough annotated training data is available, supervised deep-learning algorithms excel at estimating human body pose and shape using a single camera. The effects of too little such data being available can be mitigated by using other information sources, such as databases of body shapes, to learn priors. Unfortunately, such sources are not always available either. We show that, in such cases, easy-to-obtain unannotated videos can be used instead to provide the required supervisory signals. Given a trained model using too little annotated data, we compute poses in consecutive frames along with the optical flow between them. We then enforce consistency between the image optical flow and the one that can be inferred from the change in pose from one frame to the next. This provides enough additional supervision to effectively refine the network weights and to perform on par with methods trained using far more annotated data.

Edge applications, such as collaborative robotics and spacecraft rendezvous, demand efficient 6D object pose estimation on resource-constrained embedded platforms. Existing 6D pose estimation networks are often too large for such deployments, necessitating compression while maintaining reliable performance. To address this challenge, we introduce Modular Quantization-Aware Training (MQAT), an adaptive and mixed-precision quantization-aware training strategy that exploits the modular structure of modern 6D pose estimation architectures. MQAT guides a systematic gradated modular quantization sequence and determines module-specific bit precisions, leading to quantized models that outperform those produced by state-of-the-art uniform and mixed-precision quantization techniques. Our experiments showcase the generality of MQAT across datasets, architectures, and quantization algorithms. Remarkably, MQAT-trained quantized models achieve a significant accuracy boost (>7%) over the baseline full-precision network while reducing model size by a factor of 4x or more.

We study the problem of unsupervised heteroscedastic covariance estimation, where the goal is to learn the multivariate target distribution $\mathcal{N}(y, \Sigma_y | x )$ given an observation $x$. This problem is particularly challenging as $\Sigma_{y}$ varies for different samples (heteroscedastic) and no annotation for the covariance is available (unsupervised). Typically, state-of-the-art methods predict the mean $f_{\theta}(x)$ and covariance $\textrm{Cov}(f_{\theta}(x))$ of the target distribution through two neural networks trained using the negative log-likelihood. This raises two questions: (1) Does the predicted covariance truly capture the randomness of the predicted mean? (2) In the absence of ground-truth annotation, how can we quantify the performance of covariance estimation? We address (1) by deriving TIC: Taylor Induced Covariance, which captures the randomness of the multivariate $f_{\theta}(x)$ by incorporating its gradient and curvature around $x$ through the second order Taylor polynomial. Furthermore, we tackle (2) by introducing TAC: Task Agnostic Correlations, a metric which leverages conditioning of the normal distribution to evaluate the covariance. We verify the effectiveness of TIC through multiple experiments spanning synthetic (univariate, multivariate) and real-world datasets (UCI Regression, LSP, and MPII Human Pose Estimation). Our experiments show that TIC outperforms state-of-the-art in accurately learning the covariance, as quantified through TAC.

Prior methods that tackle the problem of generalizable object pose estimation highly rely on having dense views of the unseen object. By contrast, we address the scenario where only a single reference view of the object is available. Our goal then is to estimate the relative object pose between this reference view and a query image that depicts the object in a different pose. In this scenario, robust generalization is imperative due to the presence of unseen objects during testing and the large-scale object pose variation between the reference and the query. To this end, we present a new hypothesis-and-verification framework, in which we generate and evaluate multiple pose hypotheses, ultimately selecting the most reliable one as the relative object pose. To measure reliability, we introduce a 3D-aware verification that explicitly applies 3D transformations to the 3D object representations learned from the two input images. Our comprehensive experiments on the Objaverse, LINEMOD, and CO3D datasets evidence the superior accuracy of our approach in relative pose estimation and its robustness in large-scale pose variations, when dealing with unseen objects.

We introduce the first multitasking vision transformer adapters that learn generalizable task affinities which can be applied to novel tasks and domains. Integrated into an off-the-shelf vision transformer backbone, our adapters can simultaneously solve multiple dense vision tasks in a parameter-efficient manner, unlike existing multitasking transformers that are parametrically expensive. In contrast to concurrent methods, we do not require retraining or fine-tuning whenever a new task or domain is added. We introduce a task-adapted attention mechanism within our adapter framework that combines gradient-based task similarities with attention-based ones. The learned task affinities generalize to the following settings: zero-shot task transfer, unsupervised domain adaptation, and generalization without fine-tuning to novel domains. We demonstrate that our approach outperforms not only the existing convolutional neural network-based multitasking methods but also the vision transformer-based ones. Our project page is at \url{https://ivrl.github.io/VTAGML}.

While road obstacle detection techniques have become increasingly effective, they typically ignore the fact that, in practice, the apparent size of the obstacles decreases as their distance to the vehicle increases. In this paper, we account for this by computing a scale map encoding the apparent size of a hypothetical object at every image location. We then leverage this perspective map to (i) generate training data by injecting onto the road synthetic objects whose size corresponds to the perspective foreshortening; and (ii) incorporate perspective information in the decoding part of the detection network to guide the obstacle detector. Our results on standard benchmarks show that, together, these two strategies significantly boost the obstacle detection performance, allowing our approach to consistently outperform state-of-the-art methods in terms of instance-level obstacle detection.

Dense Self-Supervised Learning (SSL) methods address the limitations of using image-level feature representations when handling images with multiple objects. Although the dense features extracted by employing segmentation maps and bounding boxes allow networks to perform SSL for each object, we show that they suffer from coupling and positional bias, which arise from the receptive field increasing with layer depth and zero-padding. We address this by introducing three data augmentation strategies, and leveraging Figure 1: Object coupling and positional bias on the them in (i) a decoupling module that aims to robustify region-level features extracted from a scene with multiple the network to variations in the object's surroundings, and objects. The overlap between objects increases with the (ii) a de-positioning module that encourages the network to depth within the network, thus mixing information across discard positional object information.

Object location priors have been shown to be critical for the standard 6D object pose estimation setting, where the training and testing objects are the same. Specifically, they can be used to initialize the 3D object translation and facilitate 3D object rotation estimation. Unfortunately, the object detectors that are used for this purpose do not generalize to unseen objects, i.e., objects from new categories at test time. Therefore, existing 6D pose estimation methods for previously-unseen objects either assume the ground-truth object location to be known, or yield inaccurate results when it is unavailable. In this paper, we address this problem by developing a method, LocPoseNet, able to robustly learn location prior for unseen objects. Our method builds upon a template matching strategy, where we propose to distribute the reference kernels and convolve them with a query to efficiently compute multi-scale correlations. We then introduce a novel translation estimator, which decouples scale-aware and scale-robust features to predict different object location parameters. Our method outperforms existing works by a large margin on LINEMOD and GenMOP. We further construct a challenging synthetic dataset, which allows us to highlight the better robustness of our method to various noise sources.

Knowledge distillation facilitates the training of a compact student network by using a deep teacher one. While this has achieved great success in many tasks, it remains completely unstudied for image-based 6D object pose estimation. In this work, we introduce the first knowledge distillation method driven by the 6D pose estimation task. To this end, we observe that most modern 6D pose estimation frameworks output local predictions, such as sparse 2D keypoints or dense representations, and that the compact student network typically struggles to predict such local quantities precisely. Therefore, instead of imposing prediction-to-prediction supervision from the teacher to the student, we propose to distill the teacher's \emph{distribution} of local predictions into the student network, facilitating its training. Our experiments on several benchmarks show that our distillation method yields state-of-the-art results with different compact student models and for both keypoint-based and dense prediction-based architectures.