Search

Abstract We review the current geoscientific knowledge of the volcanic unrest of 2004–2005 on Tenerife (Canary Islands) and revisit its gravimetric imprint. We revise the interpretation of the observed spatiotemporal (time-lapse) gravity changes accompanying the unrest by applying the Growth inversion approach based on model exploration and free geometry growing source bodies. We interpret the Growth solution, our new gravimetric model of the unrest, in the context of structural controls and the existing volcanological and geological knowledge of the central volcanic complex (CVC) of the island. Structural controls are inferred from the updated structural subsurface CVC density model obtained by our new Growth inversion of the available complete Bouguer anomalies (CBA data). Our gravimetric picture sees the unrest as a failed eruption, due to a stalled magma intrusion in the central position below the Teide–Pico Viejo stratocones, followed by upward and lateral migration of volcanic fluids reaching the aquifer and the SW end of the caldera wall. We thus classify the volcanic unrest of 2004–2005 as hybrid, in agreement with previous studies. The Growth inversion indicates that magma propagated along the boundary between the basaltic core of the island, the Boca Tauce volcanic body and the more permeable (less compacted) volcanic rocks with lower density. This gravimetric picture of the unrest provides new insights into the potential future reactivation of the volcanic system. Article Highlights Current geoscientific knowledge of the Tenerife volcanic unrest of 2004–2005 is reviewed New insights into the unrest are yielded by Growth inversion of observed time-lapse gravity changes Role of the freely adjustable inversion parameters in the Growth methodology is demonstrated Pros and cons of the Growth inversion approach in volcano gravimetric applications are illustrated

Thin elongated sources, such as dykes, sills, chimneys, inclined sheets, etc., often encountered in volcano gravimetric studies, pose great challenges to gravity inversion methods based on model exploration and growing sources bodies. The Growth inversion approach tested here is based on partitioning the subsurface into right-rectangular cells and populating the cells with differential densities in an iterative weighted mixed adjustment process, in which the minimization of the data misfit is balanced by forcing the growing subsurface density distribution into compact source bodies. How the Growth inversion can cope with thin elongated sources is the subject of our study. We use synthetic spatiotemporal gravity changes caused by simulated sources placed in three real volcanic settings. Our case studies demonstrate the benefits and limitations of the Growth inversion as applied to sparse and noisy gravity change data generated by thin elongated sources. Such sources cannot be reproduced by Growth accurately. They are imaged with smaller density contrasts, as much thicker, with exaggerated volume. Despite this drawback, the Growth inversion can provide useful information on several source parameters even for thin elongated sources, such as the position (including depth), the orientation, the length, and the mass, which is a key factor in volcano gravimetry. Since the density contrast of a source is not determined by the inversion, but preset by the user to run the inversion process, it cannot be used to specify the nature of the source process. The interpretation must be assisted by external constraints such as structural or tectonic controls, or volcanological context. Synthetic modeling and Growth inversions, such as those presented here, can serve also for optimizing the volcano monitoring gravimetric network design. We conclude that the Growth inversion methodology may, in principle, prove useful even for the detection of thin elongated sources of high density contrast by providing useful information on their position, shape (except for thickness) and mass, despite the strong ambiguity in determining their differential density and volume. However, this yielded information may be severely compromised in reality by the sparsity and noise of the interpreted gravity data.

We report on an 11-year-old girl with cystic fibrosis who presented with thoracic pain and an extensive subcutaneous emphysema and subsequently developed progressive respiratory distress. The chest computed tomography revealed a huge pneumomediastinum. Due to the development of severe respiratory failure, urgent needle thoracocentesis was necessary that resulted in only temporary improvement. Therefore, under general anesthesia two mediastinal drains were introduced. Using active suction, the size of the pneumomediastinum decreased gradually and the drains were removed after 3 weeks. Here, we describe an extremely rare situation, when acute surgical intervention was necessary in a child with spontaneous pneumomediastinum.

There are several options available for conservative treatment of partial-thickness burns, however, reliable, affordable, and easily obtainable animal testing models are hard to find for the comparison of the different treatment methods. We aimed at developing a preclinical testing model and at comparing four treatment methods for superficial partial-thickness burns.Burn injury was induced in 90 adult male Wistar rats by placing the 130°C hot tip of a commercially obtainable soldering device for 30 s on the clipped skin of the interscapular region at a steady pressure. Skin histology was studied on days 5, 10, and 22 after the induction of the burn injury, on which days, respectively, the ratio of the not epithelialized wound (%), the extent of re-epithelialization (score), and the scar thickness (µm) were assessed. We compared 4 groups: silver-sulfadiazine cream, zinc-hyaluronan gel, silver foam dressing, and the combination of zinc-hyaluronan gel with a silver foam dressing.On day 5, the induction of superficial partial-thickness burn injury was confirmed histologically in the rats. The zinc-hyaluronan gel and the combination treatment resulted in a markedly smaller ratio of the non-epithelialized area (29 ± 10% and 28 ± 13%, respectively) than silver-sulfadiazine cream (69 ± 4%; p0.01). On day 10, the extent of re-epithelialization was the lowest (∼0.2) in the silver-sulfadiazine cream group, while the other 3 treatments performed significantly better. The combination treatment lead to the maximal score of 2 in all rats, which was higher than in the other 3 treatment groups. On day 22, the scar thickness was the smallest in the combination treatment group (560 ± 42 µm), which was significantly less than in the silver-sulfadiazine cream group (712 ± 38 µm; p0.05).We designed and histologically confirmed a reproducible method for induction of superficial partial-thickness burns in rats for preclinical testing. In our model, the combination of zinc-hyaluronan gel with silver foam dressing was more effective than either of its components alone or than silver-sulfadiazine cream.

New gravimetric observations were carried out in the summit area of Mt. Etna in July 2021. Only the north-west half of the planned survey area was accessible to field work due to ongoing intense eruptive activity. The new gravimetric observation points (171 in number) were positioned using precise geodetic positioning based on GNSS technology. Due to rough conditions, unstable ground and wind, the gravity observations were collected with a precision of about 15 microGal using two relative gravimeters (CG5 and CG6). Complete Bouguer anomalies (CBA) were compiled. The computation of accurate topographic correction for CBA compilation poses a challenge because of the ever-changing topography around the summit craters due to intense eruptive activity. Precise topographic correction was computed using the Toposk software. The available high resolution (5 m) DEM released in 2016, and the reference constant topographic density of 2300 kg/m3, which resulted from our analysis as representative for the summit area, were adopted for the numerical evaluation of the topographic correction. These data will serve the 2D and 3D density modelling for determining the subsurface structural model of the summit area and the upper-most part of magma feeders of the summit craters on Etna.

To evaluate if laparoscopic appendectomy is more favorable than open appendectomy. We included in the study paediatric appendectomy cases from a single institute performed between 2012 and 2018. Patients were divided into two groups: laparoscopic (LA) and open appendectomy (OA, including converted LA). Co-morbidities were excluded. Patients were subdivided according to the histological results as uncomplicated (UCAA—simplex, catharral, phlegmonous) and complicated acute appendicitis (CAA—gangrenous, perforated) cases. Complications (wound healing problems, intra-abdominal abscess, ileus) and length of stay (LOS) were analyzed. For statistical analysis, Fisher’s exact and Chi2 for trend tests were used. Altogether 1027 appendectomies (330 LA, 697 OA) were performed. The mean age at operation was 10.9 years (SD 3.4 years). After LA 26/330 and after OA 65/697 complications were observed. Median LOS in patients after LA was 2 days; whereas following OA, it was 4 (p

Background Appendicitis is one of the most common diagnoses in pediatric populations. Although new recommendations for the treatment of pediatric appendicitis were published, management varies among different institutions. Objectives To determine current practices in 4 (n = 4) representative pediatric surgical departments in Central Europe. Material and methods One department from each of the 4 countries was surveyed using an online questionnaire. Questions focused on preoperative, operative and postoperative practices in 2018, particularly those related to antibiotic (ATB) therapy and laparoscopy. Results A total of 519 appendectomies were performed, among which 413 (79.6%) were laparoscopic appendectomies (LAs), with a conversion rate of 5.1%. Appendectomy, as an elective procedure, was performed in 43 (8.3%) patients. One-quarter (129 patients) had complex appendicitis and 72.3% of these were operated laparoscopically. In 3 departments, ATB prophylaxis was administered, based on the decisions of the operating surgeon. One department used standard ATB prophylaxis (metronidazole). Whenever phlegmonous appendicitis was detected, ATB were administered therapeutically in 2 departments. Two other departments administered ATB based on surgeon decision. The choice of ATB was not standardized. If complex appendicitis was detected, all sites administered ATB therapeutically. The type of ATB treatment was standardized in complex cases in 2 departments. Thirty-four complications (6.6%) at surgical sites were recorded - 4.1% (16/390) after uncomplicated and 14% (18/129) after complex appendicitis. Thirty-two occurred after acute surgeries and 26 of these followed laparoscopic procedures. Postoperatively, intra-abdominal abscesses occurred in 3.5% of laparoscopic and in 2.9% of open appendectomy (OA) cases. Conclusions This questionnaire study showed that treatment outcomes for appendicitis in children in Central Europe are comparable with data reported in the literature. Laparoscopic appendectomy is the predominant surgical method, but there is a little consensus for ATB treatment in the management of appendicitis at our 4 pediatric surgical departments.

Some geophysical or geodynamic applications require the use of true vertical gradient of gravity (VGG). This demand may be associated with reductions of or corrections to observed gravity or its spatiotemporal changes. In the absence of in situ measured VGG values, the constant value of the theoretical (normal) free air gradient (FAG) is commonly used. We propose an alternative to this practice which may significantly reduce systematic errors associated with the use of constant FAG. The true VGG appears to be better approximated, in areas with prominent and rugged topography, such as alpine or some volcanic regions, by a value based on the modelled contribution of the topographic masses to the gradient. Such prediction can be carried out with a digital elevation model (DEM) of sufficient resolution and accuracy. Here we present the VGG field computed for Mt. Etna (Italy), one of the most active and best monitored volcanoes worldwide, to illustrate how strongly the VGG deviates spatially from constant FAG. The predicted (modelled) VGG field is verified by in situ observations. We also take a look at the sensitivity of the VGG prediction to the resolution and quality of used DEMs. We conclude with discussing the applicability of the topo-predicted VGG field in near surface structural and volcanological micro-gravimetric studies.

Urothelial neoplasms of the bladder (UNB) are considerably rare throughout the pediatric population. UNB develops from the urothelial tissue in the form of a benign disease, generally favoring a successful prognosis in the majority of cases. The authors present the diagnosis and treatment regarding two medical case reports in which urothelial papilloma was diagnosed and effectively treated. Case 1: A 15-year-old male patient was presented to our clinic complaining of a painless yet distinctive, macroscopic form of hematuria. Following a routine examination, which included ultrasound (US) and intravenous pyelography, the urethrocystoscopy revealed an intravesical solitary lesion positioned in the vicinity of the left ureteral orifice. Additionally, histology confirmed urothelial papilloma. During the follow-up, laboratory, urinary control tests, and US results all proved negative. Case 2: A 13-year-old male patient was admitted to our clinic and examined, in regard to complaints associated with recurrent abdominal pain. The pathology was discovered incidentally on abdominal US. Preoperative US and magnetic resonance imaging (MRI) studies ensued, resulting in a scheduled MRI, followed by urethrocystoscopy, which confirmed an intravesical solitary lesion positioned near the right ureteral orifice. Histology revealed urothelial papilloma. During the follow-up control cystoscopy, one resection was repeated due to the presence of a residual tumor. Today, 10 years since the presence of uroepithelial papilloma, both patients are asymptomatic and tumor-free. If there is likely suspicion of recurrence, cystoscopy is recommended.

Background: Laboratory skills training is an essential step before conducting minimally invasive surgery in clinical practice. Our main aim was to develop an animal model for training in clinically highly challenging laparoscopic duodenal atresia repair that could be useful in establishing a minimum number of repetitions to indicate safe performance of similar interventions on humans. Materials and methods: A rabbit model of laparoscopic duodenum atresia surgery involving a diamond-shaped duodeno-duodenostomy was designed. This approach was tested in two groups of surgeons: in a beginner group without any previous clinical laparoscopic experience (but having undergone previous standardized dry-lab training, n = 8) and in an advanced group comprising pediatric surgery fellows with previous clinical experience of laparoscopy (n = 7). Each participant performed eight interventions. Surgical time, expert assessment using the Global Operative Assessment of Laparoscopic Skills (GOALS) score, anastomosis quality (leakage) and results from participant feedback questionnaires were analyzed. Results: Participants in both groups successfully completed all eight surgeries. The surgical time gradually improved in both groups, but it was typically shorter in the advanced group than in the beginner group. The leakage rate was significantly lower in the advanced group in the first two interventions, and it reached its optimal level after five operations in both groups. The GOALS and participant feedback scores showed gradual increases, evident even after the fifth surgery. Conclusions: Our data confirm the feasibility of this advanced pediatric laparoscopic model. Surgical time, anastomosis quality, GOALS score and self-assessment parameters adequately quantify technical improvement among the participants. Anastomosis quality reaches its optimal value after the fifth operation even in novice, but uniformly trained surgeons. A minimum number of wet-lab operations can be determined before surgery can be safely conducted in a clinical setting, where the development of further non-technical skills is also required. Open access funding provided by University of Szeged. This study has been funded by FEDER funds through the Competitiveness Factors Operational Program (COMPETE) and by national funds through the Foundation for Science and Technology (FCT) within the scope of project No. POCI-01–0145-FEDER-007038, as well as by project No. NORTE-01–0145-FEDER-000013, supported by the Northern Portugal Regional Operational Program (NORTE 2020) under the Portugal 2020 Partnership Agreement through the European Regional Development Fund (FEDER).

Objectives, setting and participantsIn July 2011, the EU adopted the Falsified Medicines Directive (FMD) primarily aiming to prevent the infiltration of falsified medicines into the legitimate supply chain. Our aim was to measure the cost elements of FMD implementation and operation using an internationally adaptable tool among Hungarian hospital pharmacies.DesignA 41-item questionnaire was developed to evaluate the implementation process and associated cost consequences leading up to February 2019, and the stabilisation period.ResultsOur representative data are supported by the high response rate, as 44.8% of the Hungarian hospital pharmacies have completed the survey. Human resource requirements related to decommissioning activities were measured as working hours and were expressed in full time equivalent (FTE). We have found an increased workload with extra 0.25 pharmacist and 0.75 technician FTE/institution at the end of the stabilisation period. The entire infrastructural and IT implementation costs were determined as €2173, on average (SD: €3366) and the median was €1506 (range: €0–€20 723). The total IT investment cost per institution was valued at €1410 (SD: €335). We identified a positive correlation (R=0.663) in consideration of the costs, the number of beds and the number of hospital locations with a multivariate linear regression model. At the time of our survey, 76.7% of the respondents experienced drug shortages, 58.1% reported suspected increase in drug costs regarding serialised medications, and 53.5% noticed an increase in packaging size.ConclusionsNotably, our research is the first complex study depicting FMD cost implications in the hospital pharmacy sector in Central Europe, indicating decommissioning significantly impacted workflow referencing human resources and IT.

Traditional computer vision models are trained to predict a fixed set of predefined categories. Recently, natural language has been shown to be a broader and richer source of supervision that provides finer descriptions to visual concepts than supervised "gold" labels. Previous works, such as CLIP, use a simple pretraining task of predicting the pairings between images and text captions. CLIP, however, is data hungry and requires more than 400M image text pairs for training. We propose a data-efficient contrastive distillation method that uses soft labels to learn from noisy image-text pairs. Our model transfers knowledge from pretrained image and sentence encoders and achieves strong performance with only 3M image text pairs, 133x smaller than CLIP. Our method exceeds the previous SoTA of general zero-shot learning on ImageNet 21k+1k by 73% relatively with a ResNet50 image encoder and DeCLUTR text encoder. We also beat CLIP by 10.5% relatively on zero-shot evaluation on Google Open Images (19,958 classes)., Comment: 4 pages, 1 figure

By the one-to-many nature of the super-resolution (SR) problem, a single low-resolution (LR) image can be mapped to many high-resolution (HR) images. However, learning based SR algorithms are trained to map an LR image to the corresponding ground truth (GT) HR image in the training dataset. The training loss will increase and penalize the algorithm when the output does not exactly match the GT target, even when the outputs are mathematically valid candidates according to the SR framework. This becomes more problematic for the blind SR, as diverse unknown blur kernels exacerbate the ill-posedness of the problem. To this end, we propose a fundamentally different approach for the SR by introducing the concept of the adaptive target. The adaptive target is generated from the original GT target by a transformation to match the output of the SR network. The adaptive target provides an effective way for the SR algorithm to deal with the ill-posed nature of the SR, by providing the algorithm with the flexibility of accepting a variety of valid solutions. Experimental results show the effectiveness of our algorithm, especially for improving the perceptual quality of HR outputs.

Self-attention is a corner stone for transformer models. However, our analysis shows that self-attention in vision transformer inference is extremely sparse. When applying a sparsity constraint, our experiments on image (ImageNet- 1K) and video (Kinetics-400) understanding show we can achieve 95% sparsity on the self-attention maps while main-taining the performance drop to be less than 2 points. This motivates us to rethink the role of self-attention in vision transformer models.

Neural Architecture Search (NAS) yields state-of-the-art neural networks that outperform their best manually-designed counterparts. However, previous NAS methods search for architectures under one set of training hyper-parameters (i.e., a training recipe), overlooking superior architecture-recipe combinations. To address this, we present Neural Architecture-Recipe Search (NARS) to search both (a) architectures and (b) their corresponding training recipes, simultaneously. NARS utilizes an accuracy predictor that scores architecture and training recipes jointly, guiding both sample selection and ranking. Furthermore, to compensate for the enlarged search space, we leverage "free" architecture statistics (e.g., FLOP count) to pretrain the predictor, significantly improving its sample efficiency and prediction reliability. After training the predictor via constrained iterative optimization, we run fast evolutionary searches in just CPU minutes to generate architecturerecipe pairs for a variety of resource constraints, called FBNetV3. FBNetV3 makes up a family of state-of-the-art compact neural networks that outperform both automatically and manually-designed competitors. For example, FB-NetV3 matches both EfficientNet and ResNeSt accuracy on ImageNet with up to 2.0× and 7.1 × fewer FLOPs, respectively. Furthermore, FBNetV3 yields significant performance gains for downstream object detection tasks, improving mAP despite 18% fewer FLOPs and 34% fewer parameters than EfficientNet-based equivalents.

The Energy Community Treaty brings together nine Contracting Parties from Eastern and South East Europe, who are obliged to comply with energy-related parts of the environmental acquis, including the rules on environmental impact assessment. The Energy Community Secretariat recently assessed alleged non-compliance with these provisions in the case of a new coal-fired power plant project and initiated a dispute settlement procedure against Bosnia and Herzegovina, the Contracting Party concerned. The present article summarizes the findings of this process.

A full-term male neonate presented with a left sided cervical lump at the level of the thyroid gland. Magnetic resonance imaging (MRI) showed a benign heterogeneous solid mass with lobulated margins. The tumor underwent complete excision. Histology revealed the diagnosis of chondromesenchymal hamartoma in ectopic thyroid tissue.

We investigate the origin and implications of deformation-induced gravitational effects in interpretation of spatiotemporal gravity changes. We review the traditional approaches to handling the attraction of subsurface and surface deformations. These effects are relevant when inferring magmatic processes in volcano geodetic studies. We focus on the surface constituent, the deformation-induced topographic effect (DITE), which consists of a gradient effect called the free-air effect (FAE) and an attraction effect referred here as the topographic deformation effect. We present defining, alternate, as well as approximate expressions for evaluating the DITE. The alternate expressions shed light on the physical nature of DITE. By simulating numerically synthetic displacement fields of diverse shapes and areal extents imposed over terrain of various relief shapes in a referential volcanic area of prominent and rugged topographic relief, we assess the suitability and accuracy of the various approximations of DITE. Synthetic case studies are carried out using a high-resolution high-accuracy DEM and the Toposk software for evaluation of topographic attraction terms. We discuss the particularities and complications in numerical evaluation of each of the DITE expressions. We close with a conclusion that the best numerical prescription for accurate evaluation of DITE is Eq. (18) derived herein. Its numerical realization requires the knowledge of the deformation field in areal form. If the vertical displacements are known only at benchmarks, two approximations of DITE are at hand that can be numerically evaluated: the normal-FAE approximation (nFAE-DITE) and the planar Bouguer approximation (BCFAG-DITE). Based on synthetic simulations, we specify under what circumstances which approximation performs better.

Results from a detailed gravity survey realized along the planned highway tunnel in the karstic area of Slovak Karst in the eastern Slovakia are presented. Detailed gravity profiles crossed an area of rugged topography, therefore the terrain corrections played a crucial role in the gravity data processing. The airborne laser scanning technique (LiDAR) was used in order to compile a high-resolution digital terrain model (DTM) of the surrounding area and to calculate terrain corrections properly. The difference between the Bouguer anomalies calculated with an available nationwide DTM and those with new LiDAR-based model can be significant in some places as it is presented in the paper. A new method for Bouguer correction density analysis based on surface data is presented. Special underground gravity measurements in the existing nearby railway tunnel were also conducted in order to determine the mean density of the topographic rocks. The Bouguer anomalies were used to interpret lithological contacts and tectonic/karstic discontinuities.

The accurate deformation-induced topographic effect (DITE) should be used to account for the gravitational effect of surface deformation when analyzing residual spatiotemporal (time-lapse) gravity changes in volcano gravimetric or 4D micro-gravimetric studies, in general. Numerical realization of DITE requires the deformation field available in grid form. We compute the accurate DITE correction for gravity changes observed at the Laguna del Maule volcanic field in Chile over three nearly annual periods spanning 2013–2016 and compare it numerically with the previously used free-air effect (FAE) correction. We assess the impact of replacing the FAE by DITE on the model source parameters of analytic inversion solutions and apply a new inversion approach based on model exploration and growing source bodies. The new inversion results based on the DITE correction shift the position of the mass intrusion upwards by a few hundred meters and lower the total mass of the migrated fluids to roughly a half, compared to the inversion results based on the local-FAE correction. Our new Growth inversion results indicate that vertical dip-slip faults beneath the lake, as well as the Troncoso fault play active roles in hosting migrating liquid. We also show that for the study period, the DITE at Laguna del Maule can be accurately evaluated by the planar Bouguer approximation, which only requires the availability of elevation changes at gravity network benchmarks. We hypothesize that this finding may be generalized to all volcanic areas with flatter or less rugged terrain and may alter interpretations based on the commonly used FAE corrections.

We address the task of domain adaptation in object detection, where there is a domain gap between a domain with annotations (source) and a domain of interest without annotations (target). As an effective semi-supervised learning method, the teacher-student framework (a student model is supervised by the pseudo labels from a teacher model) has also yielded a large accuracy gain in cross-domain object detection. However, it suffers from the domain shift and generates many low-quality pseudo labels (\textit{e.g.,} false positives), which leads to sub-optimal performance. To mitigate this problem, we propose a teacher-student framework named Adaptive Teacher (AT) which leverages domain adversarial learning and weak-strong data augmentation to address the domain gap. Specifically, we employ feature-level adversarial training in the student model, allowing features derived from the source and target domains to share similar distributions. This process ensures the student model produces domain-invariant features. Furthermore, we apply weak-strong augmentation and mutual learning between the teacher model (taking data from the target domain) and the student model (taking data from both domains). This enables the teacher model to learn the knowledge from the student model without being biased to the source domain. We show that AT demonstrates superiority over existing approaches and even Oracle (fully-supervised) models by a large margin. For example, we achieve 50.9% (49.3%) mAP on Foggy Cityscape (Clipart1K), which is 9.2% (5.2%) and 8.2% (11.0%) higher than previous state-of-the-art and Oracle, respectively., Comment: 10 pages including references. Project page: https://yujheli.github.io/projects/adaptiveteacher.html

In situ values of vertical gradients of gravity (VGGs) are often needed when compiling residual spatiotemporal gravity changes that are interpreted in volcanic areas with the objective of drawing inferences about sources of volcanic unrest or pending eruptions. VGG values are seldom acquired by in situ observations. Their availability in 4D volcano-microgravimetric surveys and studies can be mediated by predicting the VGGs based on high resolution high accuracy DEMs and modelling the topographic component (constituent) of the VGG. Based on a modelling effort and in situ verification of VGG predicted on Etna in the summit craters area, on the north-east rift and on benchmarks of the monitoring network covering the volcano in a wider context, we learned that the VGG prediction can be improved by using drone-borne photogrammetry with GNSS ground control to produce a finer DEM in the closest vicinity of the VGG point (benchmark or field point) with resolution higher than the available high-resolution LiDAR-derived DEM, and using detailed modeling of gravity effect (on VGG) of anthropogenic objects such as walls and buildings adjacent to the VGG points. In this poster we present the methods used in the refined VGG prediction and the results of the verification of VGGs predicted on Etna.

Background and importance The aim of the falsified medicines directive (FMD 2011/62/EU) is to prevent the entry of illegitimate medicines into the legal supply chain. Despite its proposed benefits, the indepth evaluation of cost implications for hospital pharmacies is still lacking. Aim and objectives Our study evaluated the current practice of serialisation and the financial impact of the FMD in a representative sample of Hungarian hospitals. Material and methods Based on literature review and interviews with hospital pharmacy experts, a 41 item questionnaire was developed to evaluate the implementation process leading up to February 2019, and the stabilisation period that followed. Questions regarding institutional data, human resource requirements, infrastructural and IT developments, and authentication procedures were sent out to all (n=96) Hungarian hospital pharmacies in September 2019. Results A high response rate (n=43, 44,8%) allowed representative data evaluation of Hungarian hospitals. By the initial launch date of FMD, the average increase in pharmacist workload was 0.92 (±0.98) hours/day, and it was estimated to increase further by 1.13 (±1.65), equalling 0.25 pharmacist full time equivalents (FTE)/institution. Additionally, FMD seemed to increase technician workload significantly compared with pharmacists (p Conclusion and relevance Our results illustrated that the FMD had notable short and long term impact on hospital pharmacies. Our aim is to adapt this methodology to other EU countries and identify good practices in serialisation at an international level. References and/or acknowledgements Conflict of interest Corporate sponsored research or other substantive relationships: The project was supported by the European Union, co-financed by the European Social Fund (EFOP-3.6.1.-16-2016-00004) and by the UNKP-19 new national excellence programme of the Ministry of Human Capacities.

Differentiable Neural Architecture Search (DNAS) has demonstrated great success in designing state-of-the-art, efficient neural networks. However, DARTS-based DNAS's search space is small when compared to other search methods', since all candidate network layers must be explicitly instantiated in memory. To address this bottleneck, we propose a memory and computationally efficient DNAS variant: DMaskingNAS. This algorithm expands the search space by up to $10^{14}\times$ over conventional DNAS, supporting searches over spatial and channel dimensions that are otherwise prohibitively expensive: input resolution and number of filters. We propose a masking mechanism for feature map reuse, so that memory and computational costs stay nearly constant as the search space expands. Furthermore, we employ effective shape propagation to maximize per-FLOP or per-parameter accuracy. The searched FBNetV2s yield state-of-the-art performance when compared with all previous architectures. With up to 421$\times$ less search cost, DMaskingNAS finds models with 0.9% higher accuracy, 15% fewer FLOPs than MobileNetV3-Small; and with similar accuracy but 20% fewer FLOPs than Efficient-B0. Furthermore, our FBNetV2 outperforms MobileNetV3 by 2.6% in accuracy, with equivalent model size. FBNetV2 models are open-sourced at https://github.com/facebookresearch/mobile-vision., Comment: 8 pages, 10 figures, accepted to CVPR 2020

3D photography is a new medium that allows viewers to more fully experience a captured moment. In this work, we refer to a 3D photo as one that displays parallax induced by moving the viewpoint (as opposed to a stereo pair with a fixed viewpoint). 3D photos are static in time, like traditional photos, but are displayed with interactive parallax on mobile or desktop screens, as well as on Virtual Reality devices, where viewing it also includes stereo. We present an end-to-end system for creating and viewing 3D photos, and the algorithmic and design choices therein. Our 3D photos are captured in a single shot and processed directly on a mobile device. The method starts by estimating depth from the 2D input image using a new monocular depth estimation network that is optimized for mobile devices. It performs competitively to the state-of-the-art, but has lower latency and peak memory consumption and uses an order of magnitude fewer parameters. The resulting depth is lifted to a layered depth image, and new geometry is synthesized in parallax regions. We synthesize color texture and structures in the parallax regions as well, using an inpainting network, also optimized for mobile devices, on the LDI directly. Finally, we convert the result into a mesh-based representation that can be efficiently transmitted and rendered even on low-end devices and over poor network connections. Altogether, the processing takes just a few seconds on a mobile device, and the result can be instantly viewed and shared. We perform extensive quantitative evaluation to validate our system and compare its new components against the current state-of-the-art., Comment: Project page: https://facebookresearch.github.io/one_shot_3d_photography/ Code: https://github.com/facebookresearch/one_shot_3d_photography