Fully-supervised salient object detection (SOD) methods have made great progress, but such methods often rely on a large number of pixel-level annotations, which are time-consuming and ...labour-intensive. In this paper, we focus on a new weakly-supervised SOD task under hybrid labels, where the supervision labels include a large number of coarse labels generated by the traditional unsupervised method and a small number of real labels. To address the issues of label noise and quantity imbalance in this task, we design a new pipeline framework with three sophisticated training strategies. In terms of model framework, we decouple the task into label refinement sub-task and salient object detection sub-task, which cooperate with each other and train alternately. Specifically, the R-Net is designed as a two-stream encoder-decoder model equipped with Blender with Guidance and Aggregation Mechanisms (BGA), aiming to rectify the coarse labels for more reliable pseudo-labels, while the S-Net is a replaceable SOD network supervised by the pseudo labels generated by the current R-Net. Note that, we only need to use the trained S-Net for testing. Moreover, in order to guarantee the effectiveness and efficiency of network training, we design three training strategies, including alternate iteration mechanism, group-wise incremental mechanism, and credibility verification mechanism. Experiments on five SOD benchmarks show that our method achieves competitive performance against weakly-supervised/unsupervised methods both qualitatively and quantitatively. The code and results can be found from the link of https://rmcong.github.io/proj_Hybrid-Label-SOD.html .
Residence time distributions (RTDs) are quickly becoming an important component of the quality control strategy for continuous manufacturing (CM) of pharmaceutical products. However, methodologies ...for determining RTDs, including appropriate selection of tracers and proper processing of RTD data, have not been fully developed, and there is potential to improve current methods.
This work aims to implement protocols for tracer selection and establish a methodology for the determination and quantitative comparison of RTD curves obtained by the pulse method. We used an extensive material property data library and multivariate analyses to select appropriate tracers for active pharmaceutical ingredient (API) of interest. Tracer performance was evaluated in two continuous blenders using pulse input experiments. The equivalence of RTDs obtained from the API and the tracer was examined for formulations with widely different blend flow properties.
For the most suitable tracer candidate, material properties were determined for mixtures with different API concentrations to ensure that blend properties containing varying amounts of tracer along the RTD profile remained similar to those of the target formulation. It was observed that even small changes in the physical properties of the target formulation would result in a different tracer RTD. This observation emphasized the importance of preserving the physical properties of the blend when characterizing the dynamical behavior of a unit operation.
Further, the RTD of the tracer and the API in the blender were characterized under three different base flow properties. The outcome of these experiments enabled a fair side-by-side comparison of RTDs obtained using pulses of either API or tracer for powders with a varying range of properties. In finding an appropriate tracer, RTD profiles were compared quantitively using different methods. Multivariate analysis of variance (MANOVA) was shown to be a suitable tool for RTD profile comparisons.
Display omitted
•Implementing protocols to select a tracer for RTD characterization.•Used an extensive material property database to select tracers for the API.•Evaluated tracer performance in two continuous blenders.•Demonstrated that the base powder properties can impact RTD profiles substantially.•MANOVA was shown to be a suitable tool for RTD profile comparisons.
Powder mixing is a complex process and a critical step in production across various industries. Passive acoustic emissions provide a promising potential for inline powder mixing monitoring. ...Vibrations from these emissions were measured during rotations of a V-blender using glass beads as a model solid as well as starch granules. Vibration profiles were correlated to specific phases of particle motion within the revolving V-shells. The phase, labeled Feature #1b, associated with particle collisions of the V-shell lid with attached accelerometer provided the largest measured vibration amplitudes and most reliable and relevant information on particle movement. After wavelet filtering to remove large-scale vibrations of the V-shell, the optimal information extraction method was identified as the average of the 50 highest amplitudes within the Feature #1b phase. An example illustrating the mixing of two size ranges of granules shows the potential of this monitoring method for pharmaceutical powder mixing in V-blenders.
Display omitted
•Passive acoustic emissions are measured by a sensor on the lid of a V-blender.•Passive acoustic emissions from collisions reflect particle size and flow.•The largest emission amplitudes occur from collisions with the V-blender lids.•The 50 largest emissions amplitudes allow a mixing profile to be estimated.
The discrete element method (DEM) and experiments were used to examine the mixing mechanisms and flow patterns in a twin-paddle blender containing two types of non-spherical particles. The ...applicability of the GPU-based DEM model was demonstrated through calibration tests using a classical rotary drum. Afterward, the calibrated DEM model was utilized to investigate the impact of factors such as the vessel fill level, paddle rotational speed, and particle number ratio on mixing performance. The relation between mixing performance and the operational parameters was predicted using the Response Surface Method (RSM). An escalation in the fill level, coupled with a reduction in impeller speed, led to a rise in the overall number of particles that came into contact with one another, suggesting an increase in the compactness of the mixture. The computed Peclet numbers and diffusivity coefficients revealed that diffusion was the prevailing mixing mechanism.
Display omitted
•Mixing of a mixture of two types of non-spherical particles was examined.•GPU-based DEM model was calibrated using the dynamic angle of repose tests.•The impact of operating parameters on mixing performance was studied.•Pressure and shear stress were investigated for the particle bed.•Mixing performance was optimized using Response Surface Method.
Non-spherical particle flow plays a crucial role in many industrial processes. Experiments and discrete element method (DEM) were used to compare the mixing behavior of a twin paddle blender ...containing non-spherical particles with that comprising spherical particles. The DEM model was calibrated using experimental data from a rotary drum. The impacts of particle shape on mixing quality were examined using the calibrated DEM model by calculating the relative standard deviation (RSD), particle velocities, contact forces, coordination number, diffusivity coefficient, Peclet number, normal and shear stress profiles. The degree of homogeneity for spherical particles was better than those for non-spherical particles. The cubical particles exhibited the highest compactness of solid mixture in the mixer. Compared to spherical particles, non-spherical particles showed a higher resistance to movement. It was discovered that diffusion was the superior mixing mechanism. The shear and normal stresses varied with mixing time, peaking near the blade tips.
Display omitted
•The impact of particle shape on mixing quality in a twin paddle blender was studied.•GPU-based DEM model was calibrated using experimental data obtained from a rotary drum.•The cubical particles exhibited the highest compactness of solid mixture in the mixer.•Non-spherical particles had a higher flow resistance than spherical particles.•Shear and normal stresses changed with mixing time, peaking near the blade tips.
Introduction:
The oxygen concentration used in neonatal resuscitation has been a matter of debate with higher oxygen concentrations posing many adverse outcomes. Recent guidelines recommend use of ...blender to titrate FiO2 delivered during resuscitation. However, blender being unaffordable and unavailable at many peripheral institutions, we tried to use a low-flow flowmeter to titrate the oxygen and measure FiO2 delivered at different flow rates.
Methods:
From a central oxygen supply, oxygen flow was titrated using a low-flow flowmeter which was connected to a self-inflating bag and oximeter. Three variables were taken—volume of self-inflating bag, flow rate, and number of compressions per minute. FiO2 delivered with each variable, keeping the other two constant, was recorded.
Results:
The data obtained was analyzed by fitting the study variables into a stepwise multiple linear regression model and a linear equation was obtained. The model R square obtained suggested strong linear relationship between flow rate and FiO2 delivered. The model showed statistically significant association between flow rate and FiO2 delivery, whereas association with other variables was statistically insignificant. Discussion: Our study suggests that 76.57% of change in FiO2 is determined by change in flow rate. The major advantage of this study would be at resource poor settings where a low-flow flowmeter which is more cost effective can be used to titrate the FiO2 delivered during neonatal resuscitation.
Display omitted
The aim of this study was to produce sustained-release tablets by V-shaped blending of polymer and tablets without using solvents or heating, and we investigated the design of polymer ...particles with high coating performance by modifying the structure of the particles using sodium lauryl sulfate. Dry-latex particles of ammonioalkyl methacrylate copolymer were prepared by adding the surfactant into aqueous latex, followed by freeze drying. The resulting dry latex was mixed with tablets (1:10) using a blender and the resulting coated tablets were characterized. Tablet coating by the dry latex was promoted as the weight ratio of surfactant to polymer increased. At a surfactant ratio of 5%, deposition of the dry latex was most effective and the resulting coated tablets (annealed at 60 °C/75%RH for 6 h) exhibited sustained-release characteristics over a period of 2 h. The addition of SLS prevented coagulation of colloidal polymer in the freeze drying, resulting in a loose-structured dry latex. This latex was easily pulverized by V-shaped blending with tablets and the resulting fine particles with high adhesiveness were deposited on the tablets. However, at a surfactant ratio of 10%, the coating of dry latex decreased due to reduced adhesiveness.
La continua evolución de herramientas aplicadas en digitalización y modelado 3D, en el último tiempo, ha permitido su utilidad en diferentes campos científicos, uno de ellos la paleontología, dando ...beneficios en relación al estudio y la exposición de materiales fósiles. En el presente trabajo se exponen dos metodologías para digitalización de morfotipos fitolíticos como modelos tridimensionales a partir de imágenes en dos dimensiones utilizando software de libre acceso, denominados modelo de burbujas y modo esculpir. Ambas metodologías se consideran complementarias a la hora de generar un modelo de fitolito tridimensional. Esta propuesta resulta eficaz para lograr modelados 3D, generando no sólo una forma amigable para visualizar dichos morfotipos, sino también una manera efectiva para el conocimiento y entendimiento de la variedad de morfologías que pueden caracterizar a los silicofitolitos.