Search

Boesch, Li, and Suffel were the first to identify the existence of uniformly optimally reliable graphs (UOR graphs), graphs which maximize all-terminal reliability over all graphs with $n$ vertices and $m$ edges. The all-terminal reliability of a graph, and more generally a graph's all-terminal reliability polynomial $R(G;p)$, may both be obtained via the Tutte polynomial $T(G;x,y)$ of the graph $G$. Here we show that the UOR graphs found earlier are in fact maximum graphs for the Tutte polynomial itself, in the sense that they are maximum not just for all-terminal reliability but for a vast array of other parameters and polynomials that may be obtained from $T(G;x,y)$ as well. These parameters include, but are not limited to, enumerations of a wide variety of well-known orientations, partial orientations, and fourientations of $G$; the magnitudes of the coefficients of the chromatic and flow polynomials of $G$; and a wide variety of generating functions, such as generating functions enumerating spanning forests and spanning connected subgraphs of $G$. The maximality of all of these parameters is done in a unified way through the use of $(n,m)$ Tutte polynomial posets.

The independence polynomial of a graph $G$ evaluated at $-1$, denoted here as $I(G;-1)$, has arisen in a variety of different areas of mathematics and theoretical physics as an object of interest. Engstr\"om used discrete Morse theory to prove that $\left|I(G;-1)\right|\leq 2^{\phi(G)}$ where $\phi(G)$ is the decycling number of $G$, i.e., the minimum number of vertices needed to be deleted from $G$ so that the remaining graph is acyclic. Here, we improve Engstr\"om's bound by showing $\left|I(G;-1)\right|\leq 2^{\phi_3(G)}$ where $\phi_3(G)$ is the minimum number of vertices needed to be deleted from $G$ so that the resulting graph contains no induced cycles whose length is divisible by $3$. We also note that this bound is not just sharp but that every value in the range given by the bound is attainable by some connected graph.

With their meaningful geometry and their omnipresence in the 3D world, edges are extremely useful primitives in computer vision. 3D edges comprise of lines and curves, and methods to reconstruct them use either multi-view images or point clouds as input. State-of-the-art image-based methods first learn a 3D edge point cloud then fit 3D edges to it. The edge point cloud is obtained by learning a 3D neural implicit edge field from which the 3D edge points are sampled on a specific level set (0 or 1). However, such methods present two important drawbacks: i) it is not realistic to sample points on exact level sets due to float imprecision and training inaccuracies. Instead, they are sampled within a range of levels so the points do not lie accurately on the 3D edges and require further processing. ii) Such implicit representations are computationally expensive and require long training times. In this paper, we address these two limitations and propose a 3D edge mapping that is simpler, more efficient, and preserves accuracy. Our method learns explicitly the 3D edge points and their edge direction hence bypassing the need for point sampling. It casts a 3D edge point as the center of a 3D Gaussian and the edge direction as the principal axis of the Gaussian. Such a representation has the advantage of being not only geometrically meaningful but also compatible with the efficient training optimization defined in Gaussian Splatting. Results show that the proposed method produces edges as accurate and complete as the state-of-the-art while being an order of magnitude faster. Code is released at https://github.com/kunalchelani/EdgeGaussians.

Rapid growth in the popularity of AR/VR/MR applications and cloud-based visual localization systems has given rise to an increased focus on the privacy of user content in the localization process. This privacy concern has been further escalated by the ability of deep neural networks to recover detailed images of a scene from a sparse set of 3D or 2D points and their descriptors - the so-called inversion attacks. Research on privacy-preserving localization has therefore focused on preventing these inversion attacks on both the query image keypoints and the 3D points of the scene map. To this end, several geometry obfuscation techniques that lift points to higher-dimensional spaces, i.e., lines or planes, or that swap coordinates between points % have been proposed. In this paper, we point to a common weakness of these obfuscations that allows to recover approximations of the original point positions under the assumption of known neighborhoods. We further show that these neighborhoods can be computed by learning to identify descriptors that co-occur in neighborhoods. Extensive experiments show that our approach for point recovery is practically applicable to all existing geometric obfuscation schemes. Our results show that these schemes should not be considered privacy-preserving, even though they are claimed to be privacy-preserving. Code will be available at \url{https://github.com/kunalchelani/RecoverPointsNeighborhood}.

We investigate the phase ordering (pattern formation) of systems of two-dimensional core-shell particles using Monte-Carlo (MC) computer simulations and classical density functional theory (DFT). The particles interact via a pair potential having a hard core and a repulsive square shoulder. Our simulations show that on cooling, the liquid state structure becomes increasingly characterised by long wavelength density modulations, and on further cooling forms a variety of other phases, including clustered, striped and other patterned phases. In DFT, the hard core part of the potential is treated using either fundamental measure theory or a simple local density approximation, whereas the soft shoulder is treated using the random phase approximation. The different DFTs are bench-marked using large-scale grand-canonical-MC and Gibbs-ensemble-MC simulations, demonstrating their predictive capabilities and shortcomings. We find that having the liquid state static structure factor $S(k)$ for wavenumber $k$ is sufficient to identify the Fourier modes governing both the liquid and solid phases. This allows to identify from easier-to-obtain liquid state data the wavenumbers relevant to the periodic phases and to predict roughly where in the phase diagram these patterned phases arise., Comment: 42 pages, 9 figures

We propose a way to train deep learning based keypoint descriptors that makes them approximately equivariant for locally affine transformations of the image plane. The main idea is to use the representation theory of GL(2) to generalize the recently introduced concept of steerers from rotations to affine transformations. Affine steerers give high control over how keypoint descriptions transform under image transformations. We demonstrate the potential of using this control for image matching. Finally, we propose a way to finetune keypoint descriptors with a set of steerers on upright images and obtain state-of-the-art results on several standard benchmarks. Code will be published at github.com/georg-bn/affine-steerers., Comment: To be presented at ECCV 2024

The masses of the ground and isomeric states in $^{124,125}$Ag have been measured using the phase-imaging ion-cyclotron-resonance technique at the JYFLTRAP double Penning trap mass spectrometer. The ground states of $^{124}$Ag and $^{125}$Ag were found to be 30(250) keV and 250(430) keV less bound but 36 and 110 times more precise than in the Atomic Mass Evaluation 2020, respectively. The excitation energy of $^{124}$Ag$^{m}$, ${E_x = 188.2(25)}$ keV, was determined for the first time. The new precise mass values have been utilised to study the evolution of nuclear structure via two-neutron separation energies. The impact on the astrophysical rapid neutron capture process has been investigated via neutron-capture reaction rate calculations. The precision measurements indicate a more linear trend in two-neutron separation energies and reduce the mass-related uncertainties for the neutron-capture rate of $^{124}$Ag$(n,\gamma)^{125}$Ag by a factor of around 100. The new mass values also improve the mass of $^{123}$Pd, previously measured using $^{124}$Ag as a reference., Comment: 9 pages, 4 figures

This study evaluates the performance of Large Language Models (LLMs) as an Artificial Intelligence-based tutor for a university course. In particular, different advanced techniques are utilized, such as prompt engineering, Retrieval-Augmented-Generation (RAG), and fine-tuning. We assessed the different models and applied techniques using common similarity metrics like BLEU-4, ROUGE, and BERTScore, complemented by a small human evaluation of helpfulness and trustworthiness. Our findings indicate that RAG combined with prompt engineering significantly enhances model responses and produces better factual answers. In the context of education, RAG appears as an ideal technique as it is based on enriching the input of the model with additional information and material which usually is already present for a university course. Fine-tuning, on the other hand, can produce quite small, still strong expert models, but poses the danger of overfitting. Our study further asks how we measure performance of LLMs and how well current measurements represent correctness or relevance? We find high correlation on similarity metrics and a bias of most of these metrics towards shorter responses. Overall, our research points to both the potential and challenges of integrating LLMs in educational settings, suggesting a need for balanced training approaches and advanced evaluation frameworks., Comment: The article is an extended version of a paper presented at the International Workshop on AI in Education and Educational Research (AIEER) at ECAI-2024 (27th European Conference on Artificial Intelligence)

We investigate the two-dimensional behavior of colloidal patchy ellipsoids specifically designed to follow a two-step assembly process from the monomer state to mesoscopic liquid-crystal phases, via the formation of so-called bent-core units at the intermediate stage. Our model comprises a binary mixture of ellipses interacting via the Gay-Berne potential and decorated by surface patches, with the binary components being mirror-image variants of each other - referred to as left-handed and right-handed ellipses according to the position of their patches. The surface patches are designed so as in the first stage of the assembly the monomers form bent-cores units, i.e. V-shaped dimers with a specific bent angle. The Gay-Berne interactions, which act between the ellipses, drive the dimers to subsequently form the characteristic phase observed in bent-core liquid crystals. We numerically investigate -- by means of both Molecular Dynamics and Monte Carlo simulations -- the described two-step process: we first optimize a target bent-core unit and we then fully characterize its state diagram in temperature and density, defining the regions where the different liquid crystalline phases dominate.

In open-set semi-supervised learning (OSSL), we consider unlabeled datasets that may contain unknown classes. Existing OSSL methods often use the softmax confidence for classifying data as in-distribution (ID) or out-of-distribution (OOD). Additionally, many works for OSSL rely on ad-hoc thresholds for ID/OOD classification, without considering the statistics of the problem. We propose a new score for ID/OOD classification based on angles in feature space between data and an ID subspace. Moreover, we propose an approach to estimate the conditional distributions of scores given ID or OOD data, enabling probabilistic predictions of data being ID or OOD. These components are put together in a framework for OSSL, termed \emph{ProSub}, that is experimentally shown to reach SOTA performance on several benchmark problems. Our code is available at https://github.com/walline/prosub., Comment: ECCV2024

We consider the incomplete multi-graph matching problem, which is a generalization of the NP-hard quadratic assignment problem for matching multiple finite sets. Multi-graph matching plays a central role in computer vision, e.g., for matching images or shapes, so that a number of dedicated optimization techniques have been proposed. While the closely related NP-hard multi-dimensional assignment problem (MDAP) has been studied for decades in the operations research community, it only considers complete matchings and has a different cost structure. We bridge this gap and transfer well-known approximation algorithms for the MDAP to incomplete multi-graph matching. To this end, we revisit respective algorithms, adapt them to incomplete multi-graph matching, and propose their extended and parallelized versions. Our experimental validation shows that our new method substantially outperforms the previous state of the art in terms of objective and runtime. Our algorithm matches, for example, 29 images with more than 500 keypoints each in less than two minutes, whereas the fastest considered competitor requires at least half an hour while producing far worse results.

A direct measurement of the ground-state-to-ground-state electron-capture decay $Q$ value of $^{95}$Tc has been performed utilizing the double Penning trap mass spectrometer JYFLTRAP. The $Q$ value was determined to be 1695.92(13) keV by taking advantage of the high resolving power of the phase-imaging ion-cyclotron-resonance technique to resolve the low-lying isomeric state of $^{95}$Tc (excitation energy of 38.910(40) keV) from the ground state. The mass excess of $^{95}$Tc was measured to be $-$86015.95(18) keV/c$^2$, exhibiting a precision of about 28 times higher and in agreement with the value from the newest Atomic Mass Evaluation (AME2020). Combined with the nuclear energy-level data for the decay-daughter $^{95}$Mo, two potential ultra-low $Q$-value transitions are identified for future long-term neutrino-mass determination experiments. The atomic self-consistent many-electron Dirac--Hartree--Fock--Slater method and the nuclear shell model have been used to predict the partial half-lives and energy-release distributions for the two transitions. The dominant correction terms related to those processes are considered, including the exchange and overlap corrections, and the shake-up and shake-off effects. The normalized distribution of the released energy in the electron-capture decay of $^{95}$Tc to excited states of $^{95}$Mo is compared to that of $^{163}$Ho currently being used for electron-neutrino-mass determination., Comment: 13 pages, 5 figures

Few-shot image classification involves classifying images using very few training examples. Recent vision foundation models show excellent few-shot transfer abilities, but are large and slow at inference. Using knowledge distillation, the capabilities of high-performing but slow models can be transferred to tiny, efficient models. However, common distillation methods require a large set of unlabeled data, which is not available in the few-shot setting. To overcome this lack of data, there has been a recent interest in using synthetic data. We expand on this work by presenting a novel diffusion model inversion technique (TINT) combining the diversity of textual inversion with the specificity of null-text inversion. Using this method in a few-shot distillation pipeline leads to state-of-the-art accuracy among small student models on popular benchmarks, while being significantly faster than prior work. This allows us to push even tiny models to high accuracy using only a tiny application-specific dataset, albeit relying on extra data for pre-training. Popular few-shot benchmarks involve evaluation over a large number of episodes, which is computationally cumbersome for methods involving synthetic data generation. Therefore, we also present a theoretical analysis on how the variance of the accuracy estimator depends on the number of episodes and query examples, and use these results to lower the computational effort required for method evaluation. In addition, to further motivate the use of generative models in few-shot distillation, we demonstrate that our method performs better compared to training on real data mined from the dataset used to train the diffusion model. Source code will be made available at https://github.com/pixwse/tiny2., Comment: 21 pages (9 main pages + references and appendix)

Higher-dimensional automata, i.e., pointed labeled precubical sets, are a powerful combinatorial-topological model for concurrent systems. In this paper, we show that for every (nonempty) connected polyhedron there exists a shared-variable system such that the higher-dimensional automaton modeling the state space of the system has the homotopy type of the polyhedron., Comment: 24 pages

Among lattice configurations of densely packed hard ellipses, Monte Carlo simulations are used to identify the so-called parallel and diagonal lattices as the two favourable states. The free energies of these two states are computed for several system sizes employing the Einstein Crystal method. An accurate calculation of the free energy difference between the two states reveals the parallel lattice as the state with lowest free energy. The origin of the entropic difference between the two states is further elucidated by assessing the roles of the translational and rotational degrees of freedom.

Phase separation plays an role in determining the self-assembly of biological and soft-matter systems. In biological systems, liquid-liquid phase separation inside a cell leads to the formation of various macromolecular aggregates. The interaction among these aggregates is soft, i.e., these can significantly overlap at a small energy cost. From the computer simulation point of view, these complex macromolecular aggregates are generally modeled by the so-called soft particles. The effective interaction between two particles is defined via the generalized exponential potential (GEM-n) with n = 4. Here, using molecular dynamics simulations, we study the phase separation dynamics of a size-symmetric binary mixture of ultrasoft particles. We find that when the mixture is quenched to a lower temperature below the critical temperature, the two components spontaneously start to separate. Domains of the two components form, and the equal-time order parameter reveals that the domains grow in a power-law manner with exponent 1/3, which is consistent with the Lifshitz-Slyozov law for conserved systems. Further, the static structure factor shows a power-law decay with exponent 4 consistent with the Porod law.

Deep learning (DL) has greatly advanced audio classification, yet the field is limited by the scarcity of large-scale benchmark datasets that have propelled progress in other domains. While AudioSet aims to bridge this gap as a universal-domain dataset, its restricted accessibility and lack of diverse real-world evaluation use cases challenge its role as the primary resource. Therefore, we introduce $\texttt{BirdSet}$, a large-scale benchmark dataset for audio classification focusing on avian bioacoustics. $\texttt{BirdSet}$ surpasses AudioSet with over 6,800 recording hours ($\uparrow\!17\%$) from nearly 10,000 classes ($\uparrow\!18\times$) for training and more than 400 hours ($\uparrow\!7\times$) across eight strongly labeled evaluation datasets. It serves as a versatile resource for use cases such as multi-label classification, covariate shift or self-supervised learning. We benchmark six well-known DL models in multi-label classification across three distinct training scenarios and outline further evaluation use cases in audio classification. We host our dataset on Hugging Face for easy accessibility and offer an extensive codebase to reproduce our results., Comment: Under review

The masses of $^{84}$Br, $^{105}$Mo, $^{115,119,121}$Pd, $^{122}$Ag, $^{127,129}$In, $^{132}$Sb and their respective isomeric states have been measured with the JYFLTRAP Penning trap mass spectrometer using the phase-imaging ion-cyclotron-resonance technique. The excitation energies of the isomeric states in $^{132}$Sb and $^{119}$Pd were experimentally determined for the first time, while for $^{84}$Br, $^{115}$Pd and $^{127,129}$In, the precision of the mass values was substantially improved. In $^{105}$Mo and $^{121}$Pd there were no signs of a long-lived isomeric state. The ground-state measurements of $^{119}$Pd and $^{122}$Ag indicated that both are significantly more bound than the literature values. For $^{122}$Ag, there was no indication of a proposed third long-lived state. The results for the $N=49$ nucleus $^{84}$Br and isomers close to doubly magic $^{132}$Sn have been compared to the shell-model and the microscopic quasiparticle-phonon model calculations., Comment: 14 pages, 10 figures

Convolutional neural networks (CNNs) for image processing tend to focus on localized texture patterns, commonly referred to as texture bias. While most of the previous works in the literature focus on the task of image classification, we go beyond this and study the texture bias of CNNs in semantic segmentation. In this work, we propose to train CNNs on pre-processed images with less texture to reduce the texture bias. Therein, the challenge is to suppress image texture while preserving shape information. To this end, we utilize edge enhancing diffusion (EED), an anisotropic image diffusion method initially introduced for image compression, to create texture reduced duplicates of existing datasets. Extensive numerical studies are performed with both CNNs and vision transformer models trained on original data and EED-processed data from the Cityscapes dataset and the CARLA driving simulator. We observe strong texture-dependence of CNNs and moderate texture-dependence of transformers. Training CNNs on EED-processed images enables the models to become completely ignorant with respect to texture, demonstrating resilience with respect to texture re-introduction to any degree. Additionally we analyze the performance reduction in depth on a level of connected components in the semantic segmentation and study the influence of EED pre-processing on domain generalization as well as adversarial robustness.

Algebraic multigrid (AMG) is known to be an effective solver for many sparse symmetric positive definite (SPD) linear systems. For SPD systems, the convergence theory of AMG is well-understood in terms of the $A$-norm, but in a nonsymmetric setting, such an energy norm is non-existent. For this reason, convergence of AMG for nonsymmetric systems of equations remains an open area of research. A particular aspect missing from theory of nonsymmetric and indefinite AMG is the incorporation of general relaxation schemes. In the SPD setting, the classical form of optimal AMG interpolation provides a useful insight in determining the best possible two-grid convergence rate of a method based on an arbitrary symmetrized relaxation scheme. In this work, we discuss a generalization of the optimal AMG convergence theory targeting nonsymmetric problems, using a certain matrix-induced orthogonality of the left and right eigenvectors of a generalized eigenvalue problem relating the system matrix and relaxation operator. We show that using this generalization of the optimal convergence theory, one can obtain a measure of the spectral radius of the two grid error transfer operator that is mathematically equivalent to the derivation in the SPD setting for optimal interpolation, which instead uses norms. In addition, this generalization of the optimal AMG convergence theory can be further extended for symmetric indefinite problems, such as those arising from saddle point systems so that one can obtain a precise convergence rate of the resulting two-grid method based on optimal interpolation. We provide supporting numerical examples of the convergence theory for nonsymmetric advection-diffusion problems, two-dimensional Dirac equation motivated by $\gamma_5$-symmetry, and the mixed Darcy flow problem corresponding to a saddle point system., Comment: 20 pages, 5 figures, submitted for publication in the SIAM journal on scientific computing, copper mountain special section, iterative methods, 2024

Uncertainty estimation is an essential and heavily-studied component for the reliable application of semantic segmentation methods. While various studies exist claiming methodological advances on the one hand, and successful application on the other hand, the field is currently hampered by a gap between theory and practice leaving fundamental questions unanswered: Can data-related and model-related uncertainty really be separated in practice? Which components of an uncertainty method are essential for real-world performance? Which uncertainty method works well for which application? In this work, we link this research gap to a lack of systematic and comprehensive evaluation of uncertainty methods. Specifically, we identify three key pitfalls in current literature and present an evaluation framework that bridges the research gap by providing 1) a controlled environment for studying data ambiguities as well as distribution shifts, 2) systematic ablations of relevant method components, and 3) test-beds for the five predominant uncertainty applications: OoD-detection, active learning, failure detection, calibration, and ambiguity modeling. Empirical results on simulated as well as real-world data demonstrate how the proposed framework is able to answer the predominant questions in the field revealing for instance that 1) separation of uncertainty types works on simulated data but does not necessarily translate to real-world data, 2) aggregation of scores is a crucial but currently neglected component of uncertainty methods, 3) While ensembles are performing most robustly across the different downstream tasks and settings, test-time augmentation often constitutes a light-weight alternative. Code is at: https://github.com/IML-DKFZ/values, Comment: ICLR 2024 (oral)

The multinucleon transfer (MNT) reaction approach was successfully employed for the first time to measure the isomeric ratios (IRs) of $^{211}$Po (25/2$^+$) isomer and its (9/2$^+$) ground state at the IGISOL facility using a 945 MeV $^{136}$Xe beam impinged on $^{209}$Bi and $^{\rm nat}$Pb targets. The dominant production of isomers compared to the corresponding ground states was consistently revealed in the $\alpha$-decay spectra. Deduced IR of $^{211}$Po populated through the $^{136}$Xe+$^{\rm nat}$Pb reaction was found to enhance $\approx$1.8-times than observed for $^{136}$Xe+$^{209}$Bi. State-of-the-art Langevin-type model calculations have been utilized to estimate the spin distribution of an MNT residue. The computations qualitatively corroborate with the considerable increase in IRs of $^{211}$Po produced from $^{136}$Xe+$^{\rm nat}$Pb compared to $^{136}$Xe+$^{209}$Bi. Theoretical investigations indicate a weak influence of target spin on IRs. The enhancement of the $^{211}$Po isomer in the $^{136}$Xe+$^{\rm nat}$Pb over $^{136}$Xe+$^{209}$Bi can be attributed to the different proton ($p$)-transfer production routes. Estimations demonstrate an increment in the angular momentum transfer, favorable for isomer production, with increasing projectile energy. Comparative analysis indicates the two entrance channel parameters, projectile mass and $p$-transfer channels, strongly influencing the population of the high-spin isomer of $^{211}$Po (25/2$^+$). This is the first experimental and theoretical investigation on the IRs of nuclei produced via different channels of MNT reactions, with the latter quantitatively underestimating the former by a factor of two., Comment: 5 figures

Object detection applied to LiDAR point clouds is a relevant task in robotics, and particularly in autonomous driving. Single frame methods, predominant in the field, exploit information from individual sensor scans. Recent approaches achieve good performance, at relatively low inference time. Nevertheless, given the inherent high sparsity of LiDAR data, these methods struggle in long-range detection (e.g. 200m) which we deem to be critical in achieving safe automation. Aggregating multiple scans not only leads to a denser point cloud representation, but it also brings time-awareness to the system, and provides information about how the environment is changing. Solutions of this kind, however, are often highly problem-specific, demand careful data processing, and tend not to fulfil runtime requirements. In this context we propose TimePillars, a temporally-recurrent object detection pipeline which leverages the pillar representation of LiDAR data across time, respecting hardware integration efficiency constraints, and exploiting the diversity and long-range information of the novel Zenseact Open Dataset (ZOD). Through experimentation, we prove the benefits of having recurrency, and show how basic building blocks are enough to achieve robust and efficient results., Comment: 10 pages, 5 figures

The ability for a machine learning model to cope with differences in training and deployment conditions--e.g. in the presence of distribution shift or the generalization to new classes altogether--is crucial for real-world use cases. However, most empirical work in this area has focused on the image domain with artificial benchmarks constructed to measure individual aspects of generalization. We present BIRB, a complex benchmark centered on the retrieval of bird vocalizations from passively-recorded datasets given focal recordings from a large citizen science corpus available for training. We propose a baseline system for this collection of tasks using representation learning and a nearest-centroid search. Our thorough empirical evaluation and analysis surfaces open research directions, suggesting that BIRB fills the need for a more realistic and complex benchmark to drive progress on robustness to distribution shifts and generalization of ML models.

This paper studies a notion of directed homology for preordered spaces, called the homology digraph. We show that the homology digraph is a directed homotopy invariant and establish variants of the main results of ordinary singular homology theory for the homology digraph. In particular, we prove a K\"unneth formula, which enables one to compute the homology digraph of a product of preordered spaces from the homology digraphs of the components., Comment: 19 pages

Image keypoint descriptions that are discriminative and matchable over large changes in viewpoint are vital for 3D reconstruction. However, descriptions output by learned descriptors are typically not robust to camera rotation. While they can be made more robust by, e.g., data augmentation, this degrades performance on upright images. Another approach is test-time augmentation, which incurs a significant increase in runtime. Instead, we learn a linear transform in description space that encodes rotations of the input image. We call this linear transform a steerer since it allows us to transform the descriptions as if the image was rotated. From representation theory, we know all possible steerers for the rotation group. Steerers can be optimized (A) given a fixed descriptor, (B) jointly with a descriptor or (C) we can optimize a descriptor given a fixed steerer. We perform experiments in these three settings and obtain state-of-the-art results on the rotation invariant image matching benchmarks AIMS and Roto-360. We publish code and model weights at https://github.com/georg-bn/rotation-steerers., Comment: CVPR 2024 Camera ready

Obtaining annotations for complex computer vision tasks such as object detection is an expensive and time-intense endeavor involving a large number of human workers or expert opinions. Reducing the amount of annotations required while maintaining algorithm performance is, therefore, desirable for machine learning practitioners and has been successfully achieved by active learning algorithms. However, it is not merely the amount of annotations which influences model performance but also the annotation quality. In practice, the oracles that are queried for new annotations frequently contain significant amounts of noise. Therefore, cleansing procedures are oftentimes necessary to review and correct given labels. This process is subject to the same budget as the initial annotation itself since it requires human workers or even domain experts. Here, we propose a composite active learning framework including a label review module for deep object detection. We show that utilizing part of the annotation budget to correct the noisy annotations partially in the active dataset leads to early improvements in model performance, especially when coupled with uncertainty-based query strategies. The precision of the label error proposals has a significant influence on the measured effect of the label review. In our experiments we achieve improvements of up to 4.5 mAP points of object detection performance by incorporating label reviews at equal annotation budget.