Search

Liver-vessel segmentation is an essential task in the pre-operative planning of liver resection. State-of-the-art 2D or 3D convolution-based methods focusing on liver vessel segmentation on 2D CT cross-sectional views, which do not take into account the global liver-vessel topology. To maintain this global vessel topology, we rely on the underlying physics used in the CT reconstruction process, and apply this to liver-vessel segmentation. Concretely, we introduce the concept of top-k maximum intensity projections, which mimics the CT reconstruction by replacing the integral along each projection direction, with keeping the top-k maxima along each projection direction. We use these top-k maximum projections to condition a diffusion model and generate 3D liver-vessel trees. We evaluate our 3D liver-vessel segmentation on the 3D-ircadb-01 dataset, and achieve the highest Dice coefficient, intersection-over-union (IoU), and Sensitivity scores compared to prior work., Comment: Accepted in 2025 IEEE International Symposium on Biomedical Imaging (ISBI 2025)

Improving connectivity and completeness are the most challenging aspects of small liver vessel segmentation. It is difficult for existing methods to obtain segmented liver vessel trees simultaneously with continuous geometry and detail in small vessels. We proposed a diffusion model-based method with a multi-scale graph attention guidance to break through the bottleneck to segment the liver vessels. Experiments show that the proposed method outperforms the other state-of-the-art methods used in this study on two public datasets of 3D-ircadb-01 and LiVS. Dice coefficient and Sensitivity are improved by at least 11.67% and 24.21% on 3D-ircadb-01 dataset, and are improved by at least 3.21% and 9.11% on LiVS dataset. Connectivity is also quantitatively evaluated in this study and our method performs best. The proposed method is reliable for small liver vessel segmentation., Comment: This work has been submitted to the IEEE for possible publication

CNNs and computational models of biological vision share some fundamental principles, which opened new avenues of research. However, fruitful cross-field research is hampered by conventional CNN architectures being based on spatially and depthwise discrete representations, which cannot accommodate certain aspects of biological complexity such as continuously varying receptive field sizes and dynamics of neuronal responses. Here we propose deep continuous networks (DCNs), which combine spatially continuous filters, with the continuous depth framework of neural ODEs. This allows us to learn the spatial support of the filters during training, as well as model the continuous evolution of feature maps, linking DCNs closely to biological models. We show that DCNs are versatile and highly applicable to standard image classification and reconstruction problems, where they improve parameter and data efficiency, and allow for meta-parametrization. We illustrate the biological plausibility of the scale distributions learned by DCNs and explore their performance in a neuroscientifically inspired pattern completion task. Finally, we investigate an efficient implementation of DCNs by changing input contrast., Comment: Presented at ICML 2021

A number of computer vision deep regression approaches report improved results when adding a classification loss to the regression loss. Here, we explore why this is useful in practice and when it is beneficial. To do so, we start from precisely controlled dataset variations and data samplings and find that the effect of adding a classification loss is the most pronounced for regression with imbalanced data. We explain these empirical findings by formalizing the relation between the balanced and imbalanced regression losses. Finally, we show that our findings hold on two real imbalanced image datasets for depth estimation (NYUD2-DIR), and age estimation (IMDB-WIKI-DIR), and on the problem of imbalanced video progress prediction (Breakfast). Our main takeaway is: for a regression task, if the data sampling is imbalanced, then add a classification loss., Comment: Accepted at ICCV-2023

Activity progress prediction aims to estimate what percentage of an activity has been completed. Currently this is done with machine learning approaches, trained and evaluated on complicated and realistic video datasets. The videos in these datasets vary drastically in length and appearance. And some of the activities have unanticipated developments, making activity progression difficult to estimate. In this work, we examine the results obtained by existing progress prediction methods on these datasets. We find that current progress prediction methods seem not to extract useful visual information for the progress prediction task. Therefore, these methods fail to exceed simple frame-counting baselines. We design a precisely controlled dataset for activity progress prediction and on this synthetic dataset we show that the considered methods can make use of the visual information, when this directly relates to the progress prediction. We conclude that the progress prediction task is ill-posed on the currently used real-world datasets. Moreover, to fairly measure activity progression we advise to consider a, simple but effective, frame-counting baseline., Comment: Accepted at ICCVw-2023 (AI for Creative Video Editing and Understanding, ICCV workshop 2023)

Objects in videos are typically characterized by continuous smooth motion. We exploit continuous smooth motion in three ways. 1) Improved accuracy by using object motion as an additional source of supervision, which we obtain by anticipating object locations from a static keyframe. 2) Improved efficiency by only doing the expensive feature computations on a small subset of all frames. Because neighboring video frames are often redundant, we only compute features for a single static keyframe and predict object locations in subsequent frames. 3) Reduced annotation cost, where we only annotate the keyframe and use smooth pseudo-motion between keyframes. We demonstrate computational efficiency, annotation efficiency, and improved mean average precision compared to the state-of-the-art on four datasets: ImageNet VID, EPIC KITCHENS-55, YouTube-BoundingBoxes, and Waymo Open dataset. Our source code is available at https://github.com/L-KID/Videoobject-detection-by-location-anticipation., Comment: Accepted by ICCV 2023

The aim of this paper is twofold. On one hand the generalized Minkowski sets are defined and characterized. On the other hand, the Motzkin decomposable sets, along with their epigraphic versions are considered and characterized in new ways. Among them, the closed convex sets with one single minimal face, i.e. translated closed convex cones, along with their epigraphic counterparts are particularly studied.

The focused electron beam induced deposition (FEBID) process was used by employing a Gemini SEM with a beam characteristic of 1keV and 24pA for the deposition of pillars and line shaped deposits with heights between 9nm to 1um and widths from 5nm to 0.5um. All structures have been analyzed to their composition looking at a desired Si : O : C content of 1: 2 : 0. The C content of the structure was found to be ~over 60% for older deposits kept in air (~at room temperature) and less than 50% for younger deposits, only 12 hours old. Using a deposition of Si(OEt)4 at high rates and a deposition temperature of under 0 degC, an Si content of our structure between 10at% and 15at% (compositional percentage) was obtained. The FEBID structures have been deposited on Au(111) over an SiO2 wafer. The Au(111) was chosen as a substrate for the deposition of Si(OEt)4 due to its structural and morphological properties, with its surface granulation following a Chevron pattern, and the Au(111) defects having a higher contribution to the change in the composition of the final content of the structure with the increase in O ratio and a reduction in the shapes heights., Comment: 26 pages, 14 figures

With the use of proton-NMR and powder XRD (XRPD) studies, the suitability of specific Au FEBID precursors has been investigated to low electron energy, structure, excited states and resonances, structural crystal modifications, flexibility, and vaporization level. Uniquely designed precursor to meet the needs of focused electron beam induced deposition (FEBID) at the nanostructure level, the 4,5-dichloro 1,3-diethyl imidazolylidene trifluoromethyl gold(i) is a compound that proves its capability to create high purity structures, and its growing importance between other AuImx and AuClnB (where x, n are the number of radicals, B = CH, CH3 or Br) compounds in the radiation cancer therapy increases the efforts to design more suitable bonds in processes of SEM deposition and in gas-phase studies. The investigation done to its powder shape using the XRPD XPERT3 Panalytical diffractometer based on CoK{\alpha} lines show changes to its structure with temperature, level of vacuum, and light; the sensitivity of this compound makes it highly interesting to radiation research in particular. Used for FEBID, through its smaller number of C, H, and O atoms has lower levels of C contamination in the structures and on the surface, but it replaces these bonds with C-Cl and C-N bonds that have a lower bond breaking energy but still needing an extra purification step in the deposition process, whether is H2O, O2 or H jets., Comment: 27 pages, 15 figures

Deep learning has improved vanishing point detection in images. Yet, deep networks require expensive annotated datasets trained on costly hardware and do not generalize to even slightly different domains, and minor problem variants. Here, we address these issues by injecting deep vanishing point detection networks with prior knowledge. This prior knowledge no longer needs to be learned from data, saving valuable annotation efforts and compute, unlocking realistic few-sample scenarios, and reducing the impact of domain changes. Moreover, the interpretability of the priors allows to adapt deep networks to minor problem variations such as switching between Manhattan and non-Manhattan worlds. We seamlessly incorporate two geometric priors: (i) Hough Transform -- mapping image pixels to straight lines, and (ii) Gaussian sphere -- mapping lines to great circles whose intersections denote vanishing points. Experimentally, we ablate our choices and show comparable accuracy to existing models in the large-data setting. We validate our model's improved data efficiency, robustness to domain changes, adaptability to non-Manhattan settings., Comment: CVPR2022, code available at https://github.com/yanconglin/VanishingPoint_HoughTransform_GaussianSphere

The Ni(CO)4, Cr(CO)6, Co(CO)3NO are some of the most common precursors used for the focused electron-induced deposition. Some of the compounds, even though extensively used have high requirements when it comes to handling, being explosives, highly flammable, and with high toxicity levels, as is the case of Ni(CO)4. We are employing simulations to determine values that are hard to determine experimentally, and compare them with DFT calculations and experimental data where available. Using Quantemol-N cross-sections simulations for dissociative electron attachment (DEA) at low electron energy, 0 - 20eV, gives valuable information on the fragmentation of the molecules, using their bond dissociation energies, electron affinities, and incident electron energies. The values obtained for the cross-sections are 0.12x10-18cm2 for Ni(CO)4, 4.5x10-16cm2 for Co(CO)3NO DEA cross-sections and 4.3x10-15cm2for Cr(CO)6., Comment: 22 pages, 11 figures

We denote by ${\rm Hess}^+$ the set of all points $p\in\mathbb{R}^n$ such that the Hessian matrix $H_p(f)$ of the $C^2$-smooth function $f:\mathbb{R}^n\longrightarrow\mathbb{R}$ is positive definite. In this paper we provide a class of norm-coercive polynomial functions with large ${\rm Hess}^+$ regions, as their ${\rm Hess}^+$ complements happen to be bounded. A detailed analysis concerning the ${\rm Hess}^+$ region of a particular polynomial function along with some basic properties of its level curves, such as regularity, connectedness and convexity, is also provided. For such functions we also prove several properties, such as connectedness and convexity, of their level sets for sufficiently large levels. Apart from the mentioned source of such examples we provide some sufficient conditions on two functions $f,g:\mathbb{R}^2\longrightarrow\mathbb{R}$ with bounded ${\rm Hess}^+$ complements whose product $fg$ keeps having bounded ${\rm Hess}^+$ complement as well., Comment: 22 pages, 3 figures (one used twice)

The present paper intends to be a new study of a widely used precursor in nanostructure deposition and FEBID processes with a focus on its fragmentation at collisions with low-energy electrons. Newer developments in nanotechnology with applications to Focused Electron Beam Induced Deposition (FEBID) and Extreme Ultraviolet Lithography (EUVL), based on irradiation-induced chemistry come with advances in the size of the nanostructures at the surface and their flexibility in creating highly complex 3D structures. The deformation in the main structures of the FEBID process characterized by elongation, reduction in diameter of the main structure, and the deposition of additional thin layers around the structure on the substrate are results of the effect of the secondary electrons, olliding with energies lower than 20eV. Fe(CO)5 is one of the most used compounds in FEBID processes as it has high pressure and has been shown to provide high purity deposits (over 90%). This paper combines experiment and simulations to study electron scattering from Fe(CO)5, while experimental data on dissociative electron attachment is presented using the Velocity Slice Map Imaging (VMI) technique that combined with data collected on the CLUSTER apparatus at Comenius University, Bratislava and Quantemol-N simulations present the fragmentation pathways and channel distribution for each of the resulting negative ions at low energies. The Quantemol-N simulation package is used to study collision processes of low-energy electrons with Fe(CO)5 molecules including elastic, electronic excitation, and dissociative electron attachment (DEA) cross-sections., Comment: 19 pages, 6 figures

Many objects are naturally symmetric, and this symmetry can be exploited to infer unseen 3D properties from a single 2D image. Recently, NeRD is proposed for accurate 3D mirror plane estimation from a single image. Despite the unprecedented accuracy, it relies on large annotated datasets for training and suffers from slow inference. Here we aim to improve its data and compute efficiency. We do away with the computationally expensive 4D feature volumes and instead explicitly compute the feature correlation of the pixel correspondences across depth, thus creating a compact 3D volume. We also design multi-stage spherical convolutions to identify the optimal mirror plane on the hemisphere, whose inductive bias offers gains in data-efficiency. Experiments on both synthetic and real-world datasets show the benefit of our proposed changes for improved data efficiency and inference speed., Comment: BMVC 2022

Binary networks are extremely efficient as they use only two symbols to define the network: $\{+1,-1\}$. One can make the prior distribution of these symbols a design choice. The recent IR-Net of Qin et al. argues that imposing a Bernoulli distribution with equal priors (equal bit ratios) over the binary weights leads to maximum entropy and thus minimizes information loss. However, prior work cannot precisely control the binary weight distribution during training, and therefore cannot guarantee maximum entropy. Here, we show that quantizing using optimal transport can guarantee any bit ratio, including equal ratios. We investigate experimentally that equal bit ratios are indeed preferable and show that our method leads to optimization benefits. We show that our quantization method is effective when compared to state-of-the-art binarization methods, even when using binary weight pruning.

Frequency information lies at the base of discriminating between textures, and therefore between different objects. Classical CNN architectures limit the frequency learning through fixed filter sizes, and lack a way of explicitly controlling it. Here, we build on the structured receptive field filters with Gaussian derivative basis. Yet, rather than using predetermined derivative orders, which typically result in fixed frequency responses for the basis functions, we learn these. We show that by learning the order of the basis we can accurately learn the frequency of the filters, and hence adapt to the optimal frequencies for the underlying learning task. We investigate the well-founded mathematical formulation of fractional derivatives to adapt the filter frequencies during training. Our formulation leads to parameter savings and data efficiency when compared to the standard CNNs and the Gaussian derivative CNN filter networks that we build upon., Comment: Accepted at BMVC 2021

Current work on lane detection relies on large manually annotated datasets. We reduce the dependency on annotations by leveraging massive cheaply available unlabelled data. We propose a novel loss function exploiting geometric knowledge of lanes in Hough space, where a lane can be identified as a local maximum. By splitting lanes into separate channels, we can localize each lane via simple global max-pooling. The location of the maximum encodes the layout of a lane, while the intensity indicates the the probability of a lane being present. Maximizing the log-probability of the maximal bins helps neural networks find lanes without labels. On the CULane and TuSimple datasets, we show that the proposed Hough Transform loss improves performance significantly by learning from large amounts of unlabelled images., Comment: ICIP2021

Resolution in deep convolutional neural networks (CNNs) is typically bounded by the receptive field size through filter sizes, and subsampling layers or strided convolutions on feature maps. The optimal resolution may vary significantly depending on the dataset. Modern CNNs hard-code their resolution hyper-parameters in the network architecture which makes tuning such hyper-parameters cumbersome. We propose to do away with hard-coded resolution hyper-parameters and aim to learn the appropriate resolution from data. We use scale-space theory to obtain a self-similar parametrization of filters and make use of the N-Jet: a truncated Taylor series to approximate a filter by a learned combination of Gaussian derivative filters. The parameter sigma of the Gaussian basis controls both the amount of detail the filter encodes and the spatial extent of the filter. Since sigma is a continuous parameter, we can optimize it with respect to the loss. The proposed N-Jet layer achieves comparable performance when used in state-of-the art architectures, while learning the correct resolution in each layer automatically. We evaluate our N-Jet layer on both classification and segmentation, and we show that learning sigma is especially beneficial for inputs at multiple sizes., Comment: Preprint accepted by IEEE Transactions on Image Processing, 2021 (TIP). Link to final published article: https://ieeexplore.ieee.org/abstract/document/9552550

Not all video frames are equally informative for recognizing an action. It is computationally infeasible to train deep networks on all video frames when actions develop over hundreds of frames. A common heuristic is uniformly sampling a small number of video frames and using these to recognize the action. Instead, here we propose full video action recognition and consider all video frames. To make this computational tractable, we first cluster all frame activations along the temporal dimension based on their similarity with respect to the classification task, and then temporally aggregate the frames in the clusters into a smaller number of representations. Our method is end-to-end trainable and computationally efficient as it relies on temporally localized clustering in combination with fast Hamming distances in feature space. We evaluate on UCF101, HMDB51, Breakfast, and Something-Something V1 and V2, where we compare favorably to existing heuristic frame sampling methods., Comment: Accepted to CVPR 2021

In the family of Intelligent Transportation Systems (ITS), Multimodal Transport Systems (MMTS) have placed themselves as a mainstream transportation mean of our time as a feasible integrative transportation process. The Global Economy progressed with the help of transportation. The volume of goods and distances covered have doubled in the last ten years, so there is a high demand of an optimized transportation, fast but with low costs, saving resources but also safe, with low or zero emissions. Thus, it is important to have an overview of existing research in this field, to know what was already done and what is to be studied next. The main objective is to explore a beneficent selection of the existing research, methods and information in the field of multimodal transportation research, to identify industry needs and gaps in research and provide context for future research. The selective survey covers multimodal transport design and optimization in terms of: cost, time, and network topology. The multimodal transport theoretical aspects, context and resources are also covering various aspects. The survey's selection includes nowadays best methods and solvers for Intelligent Transportation Systems (ITS). The gap between theory and real-world applications should be further solved in order to optimize the global multimodal transportation system., Comment: 12 pages; Accepted: Informatica (ISSN 0868-4952)

The process of knowledge discovery involves nowadays a major number of techniques. Context-Aware Data Mining (CADM) and Collaborative Data Mining (CDM) are some of the recent ones. the current research proposes a new hybrid and efficient tool to design prediction models called Scenarios Platform-Collaborative & Context-Aware Data Mining (SP-CCADM). Both CADM and CDM approaches are included in the new platform in a flexible manner; SP-CCADM allows the setting and testing of multiple configurable scenarios related to data mining at once. The introduced platform was successfully tested and validated on real life scenarios, providing better results than each standalone technique-CADM and CDM. Nevertheless, SP-CCADM was validated with various machine learning algorithms-k-Nearest Neighbour (k-NN), Deep Learning (DL), Gradient Boosted Trees (GBT) and Decision Trees (DT). SP-CCADM makes a step forward when confronting complex data, properly approaching data contexts and collaboration between data. Numerical experiments and statistics illustrate in detail the potential of the proposed platform., Comment: 15 figures

Classical work on line segment detection is knowledge-based; it uses carefully designed geometric priors using either image gradients, pixel groupings, or Hough transform variants. Instead, current deep learning methods do away with all prior knowledge and replace priors by training deep networks on large manually annotated datasets. Here, we reduce the dependency on labeled data by building on the classic knowledge-based priors while using deep networks to learn features. We add line priors through a trainable Hough transform block into a deep network. Hough transform provides the prior knowledge about global line parameterizations, while the convolutional layers can learn the local gradient-like line features. On the Wireframe (ShanghaiTech) and York Urban datasets we show that adding prior knowledge improves data efficiency as line priors no longer need to be learned from data. Keywords: Hough transform; global line prior, line segment detection., Comment: ECCV 2020, code online: https://github.com/yanconglin/Deep-Hough-Transform-Line-Priors

Thanks to advancements in diagnosis and treatment, prostate cancer patients have high long-term survival rates. Currently, an important goal is to preserve quality-of-life during and after treatment. The relationship between the radiation a patient receives and the subsequent side effects he experiences is complex and difficult to model or predict. Here, we use machine learning algorithms and statistical models to explore the connection between radiation treatment and post-treatment gastro-urinary function. Since only a limited number of patient datasets are currently available, we used image flipping and curvature-based interpolation methods to generate more data in order to leverage transfer learning. Using interpolated and augmented data, we trained a convolutional autoencoder network to obtain near-optimal starting points for the weights. A convolutional neural network then analyzed the relationship between patient-reported quality-of-life and radiation. We also used analysis of variance and logistic regression to explore organ sensitivity to radiation and develop dosage thresholds for each organ region. Our findings show no connection between the bladder and quality-of-life scores. However, we found a connection between radiation applied to posterior and anterior rectal regions to changes in quality-of-life. Finally, we estimated radiation therapy dosage thresholds for each organ. Our analysis connects machine learning methods with organ sensitivity, thus providing a framework for informing cancer patient care using patient reported quality-of-life metrics.

Biological vision adopts a coarse-to-fine information processing pathway, from initial visual detection and binding of salient features of a visual scene, to the enhanced and preferential processing given relevant stimuli. On the contrary, CNNs employ a fine-to-coarse processing, moving from local, edge-detecting filters to more global ones extracting abstract representations of the input. In this paper we reverse the feature extraction part of standard bottom-up architectures and turn them upside-down: We propose top-down networks. Our proposed coarse-to-fine pathway, by blurring higher frequency information and restoring it only at later stages, offers a line of defence against adversarial attacks that introduce high frequency noise. Moreover, since we increase image resolution with depth, the high resolution of the feature map in the final convolutional layer contributes to the explainability of the network's decision making process. This favors object-driven decisions over context driven ones, and thus provides better localized class activation maps. This paper offers empirical evidence for the applicability of the top-down resolution processing to various existing architectures on multiple visual tasks., Comment: CVPR Workshop Deep Vision 2020