Incorrect Depth Research Articles

TSAR-MVS: Textureless-aware segmentation and correlative refinement guided multi-view stereo

The reconstruction of textureless areas has long been a challenging problem in MVS due to lack of reliable pixel correspondences between images. In this paper, we propose the Textureless-aware Segmentation And Correlative Refinement guided Multi-View Stereo (TSAR-MVS), a novel method that effectively tackles challenges posed by textureless areas in 3D reconstruction through filtering, refinement and segmentation. First, we implement the joint hypothesis filtering, a technique that merges a confidence estimator with a disparity discontinuity detector to eliminate incorrect depth estimations. Second, to spread the pixels with confident depth, we introduce an iterative correlation refinement strategy that leverages RANSAC to generate 3D planes based on superpixels, succeeded by a weighted median filter for broadening the influence of accurately determined pixels. Finally, we present a textureless-aware segmentation method that leverages edge detection and line detection for accurately identify large textureless regions for further depth completion. Experiments on ETH3D, Tanks & Temples and Strecha datasets demonstrate the superior performance and strong generalization capability of our proposed method.

Complications of collagen biostimulators in Brazil: Description of products, treatments, and evolution of 55 cases.

Complications of temporary and permanent fillers have been extensively studied. However, there is a lack of comparative data regarding poly-L-lactic acid (PLLA), calcium hydroxyapatite (CaHA), and polycaprolactone (PCL) known as collagen biostimulators. This study addressed the complications of collagen biostimulators concerning their diagnosis, type of product, treatment, and monitoring. An electronic questionnaire was sent to Brazilian dermatologic ultrasound experts to identify complications related to biostimulators. The type of biostimulator, location of application, number of vials injected, application plan, time between injection treatment and complication, injector profile, treatment, and prognosis were assessed. Fifty-five cases were identified, of which 49.1% were caused by PLLA-Elleva®, 23.6% by CaHA (alone or combined with hyaluronic acid), 20.0% by PLLA-Sculptra®, and 7.3% by PCL. The most affected area was the face (72.7%), with nodules being the most common clinical form (89.1%), generally occurring late (60.0%) (>1 month). Only one case was injected at an incorrect depth (musculoaponeurotic system-SMAS). Despite several treatments, including saline (45.5%), hyaluronidase (25.5%), diluted corticosteroids (23.6%), and energy-based devices (10.9%), only five cases showed complete resolution. Hyaluronidase was beneficial in complications related to fillers when there was an association of calcium hydroxyapatite with hyaluronic acid (p < 0.01). Complications from collagen biostimulators were more common on the face, typically manifesting about 1 month after treatment. These issues seemed to be related more to the properties of the products rather than inadequate technique. Furthermore, hyaluronidase demonstrated efficacy only in cases where there was an association with HA.

Feature‐enhanced representation with transformers for multi‐view stereo

AbstractMost existing multi‐view stereo (MVS) methods fail to consider global context information in the stage of feature extraction and cost aggregation. As transformers have shown remarkable performance on various vision tasks due to their ability to perceive global contextual information, this paper proposes a transformer‐based feature enhancement network (TF‐MVSNet) to facilitate feature representation learning by combining local features (both 2D and 3D) with long‐range contextual information. To reduce memory consumption of feature matching, the cross‐attention mechanism is leveraged to efficiently construct 3D cost volumes under the epipolar constraint. Additionally, a colour‐guided network is designed to refine depth maps at a coarse stage, hence reducing incorrect depth predictions at a fine stage. Extensive experiments were performed on the DTU dataset and Tanks and Temples (T&amp;T) benchmark and results are reported.

A Tightly Coupled Feature-Based Visual-Inertial Odometry With Stereo Cameras

Early works have shown that inertial measurement unit (IMU) can help visual odometry to achieve more accurate pose estimation. However, existing methods mainly focus on fusing visual and inertial information in the back end, while ignoring it in the front end. In this paper, we present a novel Feature-based Visual-Inertial Odometry for Stereo cameras, namely FSVIO, which makes full use of visual and inertial information in both the front and the back ends. Specifically, we firstly introduce an IMU-aided feature-based method in the visual processing part of the front end, in which IMU information is used to build robust descriptors for image perspective deformation caused by the camera motion. Then, in order to improve the efficiency of feature matching, we apply a fast-tracking method by predicting the position of feature points in the current frame with the help of combining stereo camera and IMU measurements. Furthermore, the 2D-2D epipolar geometry constraint and the improved Huber norm are introduced into the tightly coupled optimization of the back end, which reduces the influence of incorrect depth estimation from stereo cameras. Finally, our odometry is evaluated on both EuRoC datasets and real-world experiments. The experimental results verified the effectiveness and superiority of FSVIO.

Multistage Pixel-Visibility Learning With Cost Regularization for Multiview Stereo

Multiple-view stereo has potential applications in robotic operations and autonomous driving (unstructured environment construction, visual servo). With assisted depth information, inertial navigation systems can achieve precise navigation. It is, especially suitable for GPS failures in complex environments. Accurate depth estimation is a challenge in low-textured or occluded regions. To alleviate the inference of incorrect depth, a multi-stage pixel-visibility learning-based stereo network is presented in this paper. Its improvements are as follows: 1) a new content-adaptive cost volume aggregation mechanism based on neighboring pixel-wise visibility is designed to effectively produce more accurate and smoother depth map predictions in the object boundary. 2) global convolution block and boundary refinement block are developed to regularize its cost volume, they can learn the inherent constraints of feature matching correspondence and effectively mitigate the depth estimation uncertainty in low-textured regions. 3) a new loss function is designed to measure the uncertainty of predicted probability distribution and enhance the reliability of depth map inference. Experimental results on the indoor DTU datasets and the outdoor Tanks & Temples datasets indicate that our method can achieve superior performance and has a powerful generalization ability, which is comparable to state-of-the-art works. Note to Practitioners—Multiple-view stereo (MVS) can estimate dense 3D representations of scenes, which is widely used in autonomous driving, robotic navigation, virtual reality (VR), and augmented reality (AR). Aiming at the problem of incorrect depth inference in low-textured or occluded regions, this work proposes a novel multi-stage depth prediction method based on neighboring pixel-wise visibility. Our method cannot only achieve accurate depth estimation for robot perception but also make no concession to real-time performance. It is clear that the proposed method has good potential in 3D reconstruction, robotic navigation, and VR/AR fields to provide accurate depth estimation in real-time with limited memory consumption.

EEG response in humans for frequency-tagged anticorrelated random-dot stereograms: Increased coherency and alpha oscillations.

In humans, the presence of a neural mechanism triggered by anticorrelated random-dot stereograms have been theorized based on animal models from invasive studies, but have not been experimentally verified with the use of electroencephalography. In this study, we employed a phase-consistent, temporally modulated alternating depth stereogram stimulus, where we created anticorrelation by inverting the contrast between the eyes. We recorded the electrical response of the resulting brain oscillations of our four participants using EEG in both the correlated and anticorrelated conditions and whether they perceived depth movement. Our analysis found that the correlated stereograms elicited a strong coherency at the even harmonics of the depth alternation, and the anticorrelated stimulus created lower coherency peaks at the first harmonic of the depth alternation, even when participants did not report the depth movement to be visible. While both conditions created a diminishment of spectral power in the beta band, we found that the anticorrelated condition created increased spectral power in the alpha band. We experimentally verified the presence of a neural mechanism triggered by anticorrelated random-dot stereograms in the human brain with our coherency analysis and that it would not have been detected with the conventional spectral analysis due to the weakness of the response. We hypothesize that the decreased beta oscillations are related to either visual discomfort and visual attention to our stimulus, and that the increased alpha oscillations in the anticorrelated condition is a response to the incorrect depth information created by the stereogram.

A Simulation Study Using a Quality Cardiopulmonary Resuscitation Medical Manikin to Evaluate the Effects of Using Personal Protective Equipment on Performance of Emergency Resuscitation by Medical Students from the University of Silesia, Katowice, Poland and Non-Medical Personnel.

BackgroundDuring the Coronavirus disease 2019 (COVID-19) pandemic, personal protective equipment (PPE) is used during medical resuscitation aerosol-generating procedures (AGP). This simulation study aimed to evaluate the effects of PPE on the performance of emergency resuscitation by medical students from the University of Silesia, Katowice, Poland and non-medical personnel, and used a quality cardiopulmonary resuscitation (Q-CPR) medical manikin.Material/MethodsA simulation study was conducted using the Resusci Anne quality cardiopulmonary resuscitation (Q-CPR) medical manikin (Laerdal Medical AS, Norway). Participants were divided into 2 groups: a medical group of 50 and a non-medical group of 52, matched in pairs. Each pair performed 10 min of manual CPR with a compression-ventilation ratio of 30: 2 wearing PPE for AGP. The reference method was manual CPR wearing casual clothes along with surgical masks and latex gloves. Data about compression and ventilation were gathered using the QCPR Training application from Laerdal Medical.ResultsData analyses indicated statistically significant differences between medical students using PPE for AGP and basic protection: average rate of chest compressions (123 vs 114 per min; P=0.004), chest recoil (69 vs 93; P=0.0050, correct depth of chest compressions (86.5 vs 97; P=0.0081), quality of ventilation (85 vs 89; P=0.0041). Among non-medical personnel however, a statistically significant difference was in the quality of ventilation (69–85.5; P=0.0032).ConclusionsThe findings from this study showed that the use of PPE for AGP during CPR was associated with slower average speed of chest compressions, less chest recoil, incorrect depth of chest compressions, and lower quality of ventilation.

Blur image detection and classification using resnet-50

Blur classification is important for blind image restoration. It is difficult to detect blur in a single image without knowing any additional information. This paper uses edge detection methods and a deep learning convolutional neural network called Resnet-50 to classify blurry-type images. The Resnet model effectively reduces the gradient vanishing problem and uses connection skipping to train the network. Typically, images are subject to defocus blur and motion blur, which are caused by the incorrect depth of field and the movement of objects during capture. The dataset used here is the blur dataset from Kaggle, which consists of sharp images, images with blur, and motion blur. In this paper, edge detection methods are applied to images using Laplace, Sobel, Prewitt, and Roberts filters and derived features such as mean, variance, and maximum signal-to-noise ratio, which are used to train a classification algorithm for image classification.

Thermal optimization of a high-heat-load double-multilayer monochromator.

Finite-element analysis is used to study the thermal deformation of a multilayer mirror due to the heat load from the undulator beam at a low-emittance synchrotron source, specifically the ESRF-EBS upgrade beamline EBSL-2. The energy bandwidth of the double-multilayer monochromator is larger than that of the relevant undulator harmonic, such that a considerable portion of the heat load is reflected. Consequently, the absorbed power is non-uniformly distributed on the surface. The geometry of the multilayer substrate is optimized to minimize thermally induced slope errors. We distinguish between thermal bending with constant curvature that leads to astigmatic focusing or defocusing and residual slope errors. For the EBSL-2 system with grazing angles θ between 0.2 and 0.4°, meridional and sagittal focal lengths down to 100 m and 2000 m, respectively, are found. Whereas the thermal bending can be tuned by varying the depth of the `smart cut', it is found that the geometry has little effect on the residual slope errors. In both planes they are 0.1-0.25 µrad. In the sagittal direction, however, the effect on the beam is drastically reduced by the `foregiveness factor', sin(θ). Optimization without considering the reflected heat load yields an incorrect depth of the `smart cut'. The resulting meridional curvature in turn leads to parasitic focal lengths of the order of 100 m.

A Global Occlusion-Aware Approach to Self-Supervised Monocular Visual Odometry

Self-Supervised monocular visual odometry (VO) is often cast into a view synthesis problem based on depth and camera pose estimation. One of the key challenges is to accurately and robustly estimate depth with occlusions and moving objects in the scene. Existing methods simply detect and mask out regions of occlusions locally by several convolutional layers, and then perform only partial view synthesis in the rest of the image. However, occlusion and moving object detection is an unsolved problem itself which requires global layout information. Inaccurate detection inevitably results in incorrect depth as well as pose estimation. In this work, instead of locally detecting and masking out occlusions and moving objects, we propose to alleviate their negative effects on monocular VO implicitly but more effectively from two global perspectives. First, a multi-scale non-local attention module, consisting of both intra-stage augmented attention and cascaded across-stage attention, is proposed for robust depth estimation given occlusions, alleviating the impacts of occlusions via global attention modeling. Second, adversarial learning is introduced in view synthesis for monocular VO. Unlike existing methods that use pixel-level losses on the quality of synthesized views, we enforce the synthetic view to be indistinguishable from the real one at the scene-level. Such a global constraint again helps cope with occluded and moving regions. Extensive experiments on the KITTI dataset show that our approach achieves new state-of-the-art in both pose estimation and depth recovery.

Effect of Variable-Depth Tillage System on Energy Requirements for Tillage Operation and Productivity of Desert Soil

Tillage systems classified into two groups: conventional (CT) and conservation (CVT) tillage. The first group (CT), soil inverted and crop residues buried. (CT) has some disadvantages. Firstly, disturbing the soil unnecessarily in areas where the soil structure and condition is not required is wasteful of time and fuel. Secondly, incorrect tillage depth can cause damage to the soil structure, which can lead to the formation of a compaction layer. The second group, (CVT) defined as no-tillage with leaving at least 30% of the crop residue on the field. (CVT) improves the efficient usage of the natural resources of water and soil. However, (CVT) has some drawbacks such as not recommended if the soil has compaction problems. Recently a third group emerged, namely variable-depth tillage (VDT), or precision tillage technology optimizes soil physical properties only where the tillage needed by applying tillage at the required depth. (VDT) has been shown to reduce costs, labor, fuel consumption and energy requirements. To implement (VDT) system, it is necessary to determine and map soil penetration resistance, spatially and with depth through the soil profile. Therefore, an experiment conducted in a field to evaluate a technology to determine the tillage depth based on soil penetration resistant at different depths of soil. The field experiment area divided into five plots (no-tillage - uniform-depth tillage at 25cm tillage depth - uniform-depth tillage at 35cm tillage depth - uniform-depth tillage at 45cm tillage depth - variable-depth tillage) where, no-tillage indicated to (CVT), uniform-depth tillage indicated to (CT) and (VDT) indicated to precision tillage. The study measurements were fuel consumption rate (FCR), actual field capacity (AFC), power requirements (PR), specific energy (SE), operating costs (OC), soil penetration resistance (SPR) and sorghum yield (SY). The results showed that compaction layer occurred between the soil depths of 25cm and 35cm. Therefore, the (VDT) system was used at a tillage depth of 35cm. The average ratio from total field area that needed to till was about 47%. The results showed that (VDT) system caused a decreasing in the (FCR), (PR) and (OC) about 35%, 35% and 23%, respectively, compared to uniform-depth tillage system (UDT), while the (AFC) for VDT system increased about 21% compared to (UDT). The (SY) for (VDT) system increased about 53% compared to (UDT) at 25cm tillage depth, while the (SY) for VDT system decreased about 8% and 11% compared to (UDT) at 35cm and 45cm tillage depth, respectively.

Central Venous Access in Children: Technique and Complications

Children with complex diseases often need central venous catheter, not only for intraoperative use, but also for parenteral nutrition, multiple blood draw due to lab examination and to administer drugs that cannot be given via peripheral lines. Whereas the landmark driven vascular access was teached for years, nowadays the routine use of ultrasound based techniques can be called the gold standard. This article highlights standard locations for central venous access like cannulation of the internal jugular vein as well as novel alternatives such as the cannulation of the brachiocephalic vein. The correct insertion depth of central lines is essential to avoid serious complications. Several different formulas are available and can be used. Independent of the used formula, you have to make sure that complications due to incorrect depth of central venous line are a topic of the past. Finally, important tips and tricks to avoid failure and serious complications are discussed.

High physical activity vs. quality of the trunk position and the efficiency of core muscles among young males

Introduction: Physical activity is considered to be one of the keystones of a healthy lifestyle and an indispensable condition for correct body posture. The purpose of the study was to check the body posture of young males with above-average physical activity and how the position of the trunk changes after extending the arms forward. The frequency and intensity of back pain among males in the study were also analyzed. Material and methods: The study involved 50 young, healthy males, whose high daily physical activity was confirmed by the IPAQ questionnaire. Body height and weight were checked, and BMI was calculated. The position of the trunk in three planes was examined using the ultrasonic Zebris Pointer system. The examination was performed twice: in the habitual standing position and in the position taken from the Matthiass test: standing position with the arms in front of the trunk (90o). The data was prepared in the Statistica v13 program. Results: The male participants were characterized by an incorrect depth of physiological curvatures of the spine and asymmetry of the position of the shoulder and pelvic girdles in the frontal plane. Extending the arms forward in a standing position resulted in flatter thoracic kyphosis, rounder lumbar lordosis and tilting the trunk backwards. Periodic pain sensation in the lumbar spine was reported by approximately 60% of participants. It was a mild pain that did not require painkillers. There were no significant relationships between the quality of the trunk position and pains in the lumbar region. Conclusions: High physical activity turned out to be insufficient to ensure correct body posture and efficiency of core muscles. Postural education should be a permanent element of health promotion programs aimed at increasing the quality of life.

Adversarial Patch Attacks on Monocular Depth Estimation Networks

Thanks to the excellent learning capability of deep convolutional neural networks (CNN), monocular depth estimation using CNNs has achieved great success in recent years. However, depth estimation from a monocular image alone is essentially an ill-posed problem, and thus, it seems that this approach would have inherent vulnerabilities. To reveal this limitation, we propose a method of adversarial patch attack on monocular depth estimation. More specifically, we generate artificial patterns (adversarial patches) that can fool the target methods into estimating an incorrect depth for the regions where the patterns are placed. Our method can be implemented in the real world by physically placing the printed patterns in real scenes. We also analyze the behavior of monocular depth estimation under attacks by visualizing the activation levels of the intermediate layers and the regions potentially affected by the adversarial attack.

Using biomechanics to investigate the effect of VR on eye vergence system

Vergence-accommodation conflict (VAC) is the main contributor to visual fatigue during immersion in virtual environments. Many studies have investigated the effects of VAC using 3D displays and expensive complex apparatus and setup to create natural and conflicting viewing conditions. However, a limited number of studies targeted virtual environments simulated using modern consumer-grade VR headsets. Our main objective, in this work, is to test how the modern VR headsets (VR simulated depth) could affect our vergence system, in addition to investigating the effect of the simulated depth on the eye-gaze performance. The virtual scenario used included a common virtual object (a cube) in a simple virtual environment with no constraints placed on the head and neck movement of the subjects. We used ocular biomechanics and eye tracking to compare between vergence angles in matching (ideal) and conflicting (real) viewing conditions. Real vergence angle during immersion was significantly higher than ideal vergence angle and exhibited higher variability which leads to incorrect depth cues that affects depth perception and also leads to visual fatigue for prolonged virtual experiences. Additionally, we found that as the simulated depth increases, the ability of users to manipulate virtual objects with their eyes decreases, thus, decreasing the possibilities of interaction through eye gaze. The biomechanics model used here can be further extended to study muscular activity of eye muscles during immersion. It presents an efficient and flexible assessment tool for virtual environments.

Gradient‐Guided Local Disparity Editing

AbstractStereoscopic 3D technology gives visual content creators a new dimension of design when creating images and movies. While useful for conveying emotion, laying emphasis on certain parts of the scene, or guiding the viewer's attention, editing stereo content is a challenging task. Not respecting comfort zones or adding incorrect depth cues, for example depth inversion, leads to a poor viewing experience. In this paper, we present a solution for editing stereoscopic content that allows an artist to impose disparity constraints and removes resulting depth conflicts using an optimization scheme. Using our approach, an artist only needs to focus on important high‐level indications that are automatically made consistent with the entire scene while avoiding contradictory depth cues and respecting viewer comfort.

DeepToF

Time-of-flight (ToF) imaging has become a widespread technique for depth estimation, allowing affordable off-the-shelf cameras to provide depth maps in real time. However, multipath interference (MPI) resulting from indirect illumination significantly degrades the captured depth. Most previous works have tried to solve this problem by means of complex hardware modifications or costly computations. In this work, we avoid these approaches and propose a new technique to correct errors in depth caused by MPI, which requires no camera modifications and takes just 10 milliseconds per frame. Our observations about the nature of MPI suggest that most of its information is available in image space; this allows us to formulate the depth imaging process as a spatially-varying convolution and use a convolutional neural network to correct MPI errors. Since the input and output data present similar structure, we base our network on an autoencoder, which we train in two stages. First, we use the encoder (convolution filters) to learn a suitable basis to represent MPI-corrupted depth images; then, we train the decoder (deconvolution filters) to correct depth from synthetic scenes, generated by using a physically-based, time-resolved renderer. This approach allows us to tackle a key problem in ToF, the lack of ground-truth data, by using a large-scale captured training set with MPI-corrupted depth to train the encoder, and a smaller synthetic training set with ground truth depth to train the decoder stage of the network. We demonstrate and validate our method on both synthetic and real complex scenarios, using an off-the-shelf ToF camera, and with only the captured, incorrect depth as input.

Hybrid profilometry using a single monochromatic multi-frequency pattern.

In this paper, we propose a novel profilometry scheme to acquire high quality depth, where only a single shot of a monochromatic pattern is utilized. We design a band-wise pattern consisting of fringe bands spatially modulated with coprime periods. After that, with the designed pattern, depth is obtained in a hybrid manner, where both phase-based profilometry and active stereo are incorporated. To be specific, pixels in smooth regions obtain their depth values through phases analysis. Especially, based on depth smooth property, we propose a novel phase unwrapping algorithm, which avoids the problem of error propagation and yields accurate unwrapping phases. On the other hand, for boundary regions, spatial stereo, which is more robust to depth discontinuities, is utilized to modify incorrect depth values. Both theoretical verification and experimental results demonstrate that the proposed scheme can generate high quality depth maps, even for complex scenes and isolated objects.

Large aperture focus stacking with max-gradient flow by anchored rolling filtering.

Focus stacking is a computational technique to extend the depth of field through combining multiple images taken at various focus distances. However, in the large aperture case, there are always defects caused by the large blur scale, which, to the best of our knowledge, has not been well studied. In our work, we propose a max-gradient flow-based method to reduce artifacts and obtain a high-quality all-in-focus image by anchored rolling filtering. First, we define a max-gradient flow to describe the gradient propagation in the stack. The points are divided into trivial and source points with this flow. The source points are extracted as true edge points and are utilized as anchors to refine the depth map and the composited all-in-focus image iteratively. The experiments show that our method can effectively suppress the incorrect depth estimations and give a high-quality all-in-focus image.

Ultrasound-guided botulinum toxin injections

One of the key conditions for achieving the desirable result during botulinum toxin therapy for muscular dystonia, spasticity, and other diseases accompanied by spasm, pain, and autonomic dysfunction (dystonias, spasticity, etc.) is the proper administration of the agent into the muscles directly involved in the pathological process. The exact entry of botulinum toxin into the target muscles is essential for successful and safe treatment because its injection into a normal muscle may cause side effects. The most common errors are the incorrect depth and incorrect direction of a needle on insertion. Therefore, the exact injection of the agent particularly into the shallow and deep muscles is a difficult task even for an experienced specialist and requires the use of controlling methods. The European Consensus on Botulinum Toxin Therapy points out that various injection techniques are needed for the better identification of necessary muscles. However, there are currently no reports on the clear advantage of any technique. In our country, injections using palpation and anatomical landmarks have been widely used in routine practice so far; electromyographic monitoring and electrostimulation have been less frequently applied. In recent years, the new method ultrasound-guided injection has continued to grow more popular. This effective, accessible, and easy-to-use method makes it possible to manage a real-time injection process and to ensure the exact entry of the agent into the muscle. This paper is dedicated to a comparative analysis of different injection methods and to a description of the ultrasound-guided technique and its advantages over others.