RESUMO
Machine learning and computer vision have proven to be valuable tools for farmers to streamline their resource utilization to lead to more sustainable and efficient agricultural production. These techniques have been applied to strawberry cultivation in the past with limited success. To build on this past work, in this study, two separate sets of strawberry images, along with their associated diseases, were collected and subjected to resizing and augmentation. Subsequently, a combined dataset consisting of nine classes was utilized to fine-tune three distinct pretrained models: vision transformer (ViT), MobileNetV2, and ResNet18. To address the imbalanced class distribution in the dataset, each class was assigned weights to ensure nearly equal impact during the training process. To enhance the outcomes, new images were generated by removing backgrounds, reducing noise, and flipping them. The performances of ViT, MobileNetV2, and ResNet18 were compared after being selected. Customization specific to the task was applied to all three algorithms, and their performances were assessed. Throughout this experiment, none of the layers were frozen, ensuring all layers remained active during training. Attention heads were incorporated into the first five and last five layers of MobileNetV2 and ResNet18, while the architecture of ViT was modified. The results indicated accuracy factors of 98.4%, 98.1%, and 97.9% for ViT, MobileNetV2, and ResNet18, respectively. Despite the data being imbalanced, the precision, which indicates the proportion of correctly identified positive instances among all predicted positive instances, approached nearly 99% with the ViT. MobileNetV2 and ResNet18 demonstrated similar results. Overall, the analysis revealed that the vision transformer model exhibited superior performance in strawberry ripeness and disease classification. The inclusion of attention heads in the early layers of ResNet18 and MobileNet18, along with the inherent attention mechanism in ViT, improved the accuracy of image identification. These findings offer the potential for farmers to enhance strawberry cultivation through passive camera monitoring alone, promoting the health and well-being of the population.
RESUMO
Reducing meat consumption reduces carbon emissions and other environmental harms. Unfortunately, commercial plant-based meat substitutes have not seen widespread adoption. In order to enable more flexible processing methods, this paper analyzes the characteristics of commercially available spirulina, soy, pea, and brown rice protein isolates to provide data for nonmeat protein processing that can lead to cost reductions. The thermal and rheological properties, as well as viscosity, density, and particle size distribution, were analyzed for further study into alternative protein-based food processing. The differential scanning calorimetry analysis produced dry amorphous-shaped curves and paste curves with a more distinct endothermic peak. The extracted linear temperature ranges for processing within food production were 70-90 °C for spirulina, 87-116 °C for soy protein, 67-77 °C for pea protein, and 87-97 °C for brown rice protein. The viscosity analysis determined that each protein material was shear-thinning and that viscosity increased with decreased water concentration, with rice being an exception to the latter trend. The obtained viscosity range for spirulina was 15,100-78,000 cP, 3200-80,000 cP for soy protein, 1400-32,700 cP for pea protein, and 600-3500 cP for brown rice protein. The results indicate that extrusion is a viable method for the further processing of protein isolates, as this technique has a large temperature operating range and variable screw speed. The data provided here can be used to make single or multi-component protein substitutes.
RESUMO
As additive manufacturing rapidly expands the number of materials including waste plastics and composites, there is an urgent need to reduce the experimental time needed to identify optimized printing parameters for novel materials. Computational intelligence (CI) in general and particle swarm optimization (PSO) algorithms in particular have been shown to accelerate finding optimal printing parameters. Unfortunately, the implementation of CI has been prohibitively complex for noncomputer scientists. To overcome these limitations, this article develops, tests, and validates PSO Experimenter, an easy-to-use open-source platform based around the PSO algorithm and applies it to optimizing recycled materials. Specifically, PSO Experimenter is used to find optimal printing parameters for a relatively unexplored potential distributed recycling and additive manufacturing (DRAM) material that is widely available: low-density polyethylene (LDPE). LDPE has been used to make filament, but in this study for the first time it was used in the open source fused particle fabrication/fused granular fabrication system. PSO Experimenter successfully identified functional printing parameters for this challenging-to-print waste plastic. The results indicate that PSO Experimenter can provide 97% reduction in research time for 3D printing parameter optimization. It is concluded that the PSO Experimenter is a user-friendly and effective free software for finding ideal parameters for the burgeoning challenge of DRAM as well as a wide range of other fields and processes.