Transfer learning between crop types for semantic segmentation of crops versus weeds in precision agriculture

Agricultural robots rely on semantic segmentation for distinguishing between crops and weeds to perform selective treatments and increase yield and crop health while reducing the amount of chemicals used. Deep‐learning approaches have recently achieved both excellent classification performance and real‐time execution. However, these techniques also rely on a large amount of training data, requiring a substantial labeling effort, both of which are scarce in precision agriculture. Additional design efforts are required to achieve commercially viable performance levels under varying environmental conditions and crop growth stages. In this paper, we explore the role of knowledge transfer between deep‐learning‐based classifiers for different crop types, with the goal of reducing the retraining time and labeling efforts required for a new crop. We examine the classification performance on three datasets with different crop types and containing a variety of weeds and compare the performance and retraining efforts required when using data labeled at pixel level with partially labeled data obtained through a less time‐consuming procedure of annotating the segmentation output. We show that transfer learning between different crop types is possible and reduces training times for up to 80%. Furthermore, we show that even when the data used for retraining are imperfectly annotated, the classification performance is within 2% of that of networks trained with laboriously annotated pixel‐precision data.     

Friction stir welding of API X70 steel incorporating Ti-dioxide

In the present work, the microstructural evolution and hardness of HSLA X70 joined by friction stir welding were investigated. The FSW was applied to HSLA X70 with and without the addition of titanium dioxide (TiO₂) powders. To evaluate the microstructural features and hardness of different weld zones, optical microscopy and Vickers microhardness measurements were applied. The results show that the distribution of TiO₂ powders is strongly dependent on the applied friction stir processing, which in turn changed significantly the microstructure and hardness profile. In this regard, the optimum stirring action resulted in a homogeneous and fine dispersion of particles leading to the domination of an acicular ferrite phase with a hardness of 370 HV. On the other hand, the lower stirring action resulted in coarse particles as well as the development of the polygonal ferrite structure with a hardness of ～185 HV.     

Observations of Membrane Domain Reorganization in Mechanically Compressed Artificial Cells

Giant unilamellar vesicles (GUVs) are considered to be the gold standard for assembling artificial cells from the bottom up. In this study, we investigated the behavior of such biomimetic vesicles as they were subjected to mechanical compression. A microfluidic device is presented that comprises a trap to capture GUVs and a microstamp that is deflected downwards to mechanically compress the trapped vesicle. After characterization of the device, we show that single-phase GUVs can be controllably compressed to a high degree of deformation (D=0.40) depending on the pressure applied to the microstamp. A permeation assay was implemented to show that vesicle bursting is prevented by water efflux. Next, we mechanically compressed GUVs with co-existing liquid-ordered and liquid-disordered membrane phases. Upon compression, we observed that the normally stable lipid domains reorganized themselves across the surface and fused into larger domains. This phenomenon, observed here in a model membrane system, not only gives us insights into how the multicomponent membranes of artificial cells behave, but might also have interesting consequences for the role of lipid rafts in biological cells that are subjected to compressive forces in a natural environment.     

Multivariate process trajectories for molecular description of MF thermal curing and correlation with hydrolytic stability

During curing of thermosetting resins the technologically relevant properties of binders and coatings develop. However, curing is difficult to monitor due to the multitude of chemical and physical processes taking place. Precise prediction of specific technological properties based on molecular properties is very difficult. In this study, the potential of principal component analysis (PCA) and principal component regression (PCR) in the analysis of Fourier transform infrared (FTIR) spectra is demonstrated using the example of melamine-formaldehyde (MF) resin curing in solid state. FTIR/PCA-based reaction trajectories are used to visualize the influence of temperature on isothermal cure. An FTIR/PCR model for predicting the hydrolysis resistance of cured MF resin from their spectral fingerprints is presented which illustrates the advantages of FTIR/PCR compared to the combination differential scanning calorimetry/isoconversional kinetic analysis. The presented methodology is transferable to the curing reactions of any thermosetting resin and can be applied to model other technologically relevant final properties as well.     

Discovery of the Tiancilactone Antibiotics by Genome Mining of Atypical Bacterial Type II Diterpene Synthases

Although genome mining has advanced the identification, discovery, and study of microbial natural products, the discovery of bacterial diterpenoids continues to lag behind. Herein, we report the identification of 66 putative producers of novel bacterial diterpenoids, and the discovery of the tiancilactone (TNL) family of antibiotics, by genome mining of type II diterpene synthases that do not possess the canonical DXDD motif. The TNLs, which are broad‐spectrum antibiotics with moderate activities, are produced by both Streptomyces sp. CB03234 and Streptomyces sp. CB03238 and feature a highly functionalized diterpenoid skeleton that is further decorated with chloroanthranilate and γ‐butyrolactone moieties. Genetic manipulation of the tnl gene cluster resulted in TNL congeners, which provided insights into their biosynthesis and structure–activity relationships. This work highlights the biosynthetic potential that bacteria possess to produce diterpenoids and should inspire continued efforts to discover terpenoid natural products from bacteria.     

Synthesis and Characterization of Telmisartan-Derived Cell Death Modulators to Circumvent Imatinib Resistance in Chronic Myeloid Leukemia

New strategies to eradicate cancer stem cells in chronic myeloid leukemia (CML) include a combination of imatinib with peroxisome proliferator-activated receptor gamma (PPARγ) ligands. Recently, we identified the partial PPARγ agonist telmisartan as effective sensitizer of resistant K562 CML cells to imatinib treatment. Here, the importance of the heterocyclic core on the cell death-modulating effects of the telmisartan-derived lead 4′-((2-propyl-1H-benzo[d]imidazol-1-yl)methyl)-[1,1′-biphenyl]-2-carboxylic acid ( 3 b ) was investigated. Inspired by the pharmacodynamics of HYL-6d and the selective PPARγ ligand VSP-51, the benzimidazole was replaced by a carbazole or an indole core. The results indicate no correlation between PPARγ activation and sensitization of resistant CML cells to imatinib. The 2-COOH derivatives of the carbazoles or indoles achieved low activity at PPARγ, while the benzimidazoles showed 60-100 % activation. Among the 2-CO₂CH₃ derivatives, only the ester of the lead ( 2 b ) slightly activated PPARγ. Sensitizing effects were further observed for this non-cytotoxic 2 b (80 % cell death), and to a lesser extent for the lead 3 b or the 5-Br-substituted ester of the benzimidazoles ( 5 b ).     

Graphite Oxide and Aromatic Amines: Size Matters

Experimental and theoretical studies are performed in order to illuminate, for first time, the intercalation mechanism of polycyclic aromatic molecules into graphite oxide. Two representative molecules of this family, aniline and naphthalene amine are investigated. After intercalation, aniline molecules prefer to covalently connect to the graphene oxide matrix via chemical grafting, while napthalene amine molecules bind with the graphene oxide surface through π–π interactions. The presence of intercalated aromatic molecules between the graphene oxide layers is demonstrated by X‐ray diffraction, while the type of interaction between graphene oxide and polycyclic organic molecules is elucidated by X‐ray photoelectron spectroscopy. Combined quantum mechanical and molecular mechanical calculations describe the intercalation mechanism and the aniline grafting, rationalizing the experimental data. The present work opens new perspectives for the interaction of various aromatic molecules with graphite oxide and the so‐called “intercalation chemistry”.     

Effects of skim milk concentrate dry matter and spray drying air temperature on formation of capsules with varying particle size and the survival microbial cultures in a microcapsule matrix

The impact of total solid (TS) content in combination with the feed rate and air inlet temperature on the survival of Lactobacillus paracasei ssp. paracasei F19 after spray drying in a skim milk matrix has been investigated and correlated with the capsule size. Depending on the experimental conditions, the survival rates ranged from 64 to 0.2%. The higher the air inlet temperature, the lower was the survival rate and an inversely correlation between the TS content and particle size has been determined. These results clearly indicate that process stress analyses and product-related characteristics must not be regarded separately.     

Decreasing Filter Cake Resistance by Using Packing Structures

Packed beds used in absorption columns are evaluated to determine whether they can also be beneficial for cake‐forming filtrations. To assess this, model systems are characterized and separated by using a dead‐end filter cell. Filtrations are conducted with different packings; the filtrate amount over time and resulting turbidity are evaluated. Packings increase the filter cake resistance and the separation time of the cakes formed with approximately incompressible solids. However, they exhibit a positive effect on the filtration of a more complex, compressible substance; the process is not only accelerated, but also the quality of the obtained filtrate is not compromised. These results demonstrate potential in the use of packed beds for the filtration of complex biogenic suspensions.