Both these aspects make it more difficult for the model to memorize mappings while also encouraging the model to learn general patterns. Data augmentation can be a good substitute when resources are constrained” because it artificially creates more of your data when it is not possible to get more data. In the case of this project, the function being used to perform the data augmentations is set transfrom from the Hugging Face Datasets package in Python. This function performs data transformations only when the model training begins. Therefore, transformations can be done on the fly and save on computational resources. Then at each epoch, the transformations are applied to every image given to the model, so the amount of training data stays constant, but variation is added to the original data through transformations. This does not increase the number of training images as other data augmentation packages would, this artificially augments with transformations and variation. Data augmentation is an important step when training machine learning models because they can perform very powerfully if the given data sets for training are too small to train with. These models can start to over-fit, which is a problem because then the model will memorize mappings between the inputs and expected outputs. There are 54,306 images in this data set, large plastic garden pots which may seem like a lot of images, but for a machine learning model, it is not that much.

That is why data augmentation is being implemented as a step to reduce possible model over fitting.The overall performance of the pre-trained ViT image classification model with data augmentation shows good promise. The best epoch provided that F1, Accuracy, Precision, and Recall, were all equal to 1.0. This is not the best situation but shows there is room for improvement. Given the previous epochs from 1 to 8, has values for their F1 Score and Accuracy. It shows that there is possible room to improve the model with more fine-tuning and possibly an early stop in training to not have the model over-train to get the result of 1.0 for F1 Score and Accuracy. Another possible reason for the result could be the large imbalance of images between classification labels and that the Plant Village data set is considered to be a small data set. These types of results are good but not realistic for model prediction power. Even though these values are incredibly high, the training and validation loss is still fairly low and slowly converging to 0. This means that the model is learning over time, which is a good sign and shows room for improvement. A possible future improvement for this project would be to find another data set with more specific plant disease information to train the model on. Efforts will be made to look for data that includes more plant pests. That variation would be beneficial because the data set for this project mainly contains crop disease images. The disease Spider Mite, is labeled as a disease in this data set but in reality, is not a disease. It is the only classification label that has pest-inflected crop images in this data set. Having more data on crop pests would be beneficial because crop pests also cause a significant amount of crop loss and damage as well.

Technical improvements for this project include developing a stronger data augmentation technique. Instead of using the set transform method which artificially augments the data set, using another package that can actually create the separate images and add them to the data set instead would be interesting to see how it would perform. More Hyper-parameter fine-tuning could be done such as exploring the learning rate and adding an optimizer. More Epochs should be tested in order to see how the model will perform over a greater period of time. Fewer epochs will also be tested to see how early-stopping the model from training will perform. Another idea is to explore more the Training-Validation-Testing splits chosen for the data. Exploring the performance between different splits could show how to better improve the classification model. To achieve our expected agricultural need of feeding 10 billion people by 2050, we must prioritize minimizing crop loss wherever possible. More research is needed to help develop more tools to assist crop growers with preventing crop loss. The study “Mobile phone use is associated with higher smallholder agricultural productivity in Tanzania, East Africa” by Amy Quandt et al. looks into the relationship crop growers have with their cell phones as agricultural tools to help increase crop yields. “A key result is the positive association between phone use for agricultural activities and self-reported agricultural yields”. Cell phones are increasing accessibility to technological tools that help with agriculture. These technologies for assisting with crop loss will not only be utilized by commercial crop growers or the average hobbyist and enthusiast as well.

Whether crop growers use a ViT image classification model or a convolutional neural network image classification model, or another type of machine learning architecture is used, more research is needed to help develop tools to assist crop growers worldwide in eradicating crop loss everywhere!Plant–microbe associations vary widely in community composition, ecological relationships formed and microbial abundance with consequences for plant phenology and fitness . Community surveys link variation in microbial community composition in plant tissues to microbial effects on plant traits, such as maximum photosynthesis rate or abiotic stress tolerance , and experimental studies confirm these trends . Numerous factors contribute to observed variation in microbial community composition among plants, including neighbourhood effects, microbe immigration and dispersal and microbe–microbe interactions . Yet, surveys that characterise plant microbial communities cannot disentangle the simultaneous effects of dispersal, species interactions and plant traits on microbial assembly . Experimental manipulations using synthetic microbe communities have increased our knowledge of the role of plant traits and microbe arrival order in determining microbial community composition . However, synthetic community studies are often limited to one or a few model plant species. Increasing the phylogenetic breadth of experimentally inoculated plants may shed light on determinants of microbe community assembly, paving the way for inoculation strategies in agriculture that target crop growth, pathogen resistance or stress tolerance . The composition of plant-associated microbe communities can also be affected by variation in geographic location and climatic conditions which impact plant physiology, microbe species pools, growth dynamics and microbe–microbe interactions. However, studies assessing how climate variation affects the microbiome of aboveground plant tissues remain scarce . Uncoupling geographical and phenological variation from climate effects across plant species complicates efforts to examine how changing temperatures impact plant–microbe interactions on large scales . Leveraging synthetic community inoculations over seasons and changes in temperatures would allow for insights on the effects of climate and plant traits on microbial establishment and growth, particularly for plant communities in thermally variable habitats. The microbiome of flower nectar has proven a tractable model system for addressing questions in microbial ecology and plant– microbe interactions due to characteristically species-poor, highly filtered and short-lived microbial communities . Flowers mediate plant reproduction via interactions with flower-visiting animals , which also disperse microbes to nectar, leading to complex three-way interactions . Pollinators, as well as abiotic forces like wind, introduce microbes, raspberry plant pot including plant and pollinator pathogens , to flowers which may then subsequently affect floral traits or pollinator preference and plant fitness . Nectar microbiomes vary in composition across plant species , but most previous work characterises microbiomes of open flowers visited by pollinators . Pollinators introduce characteristic microbial assemblages , and consequently, studies cannot directly compare how plant species themselves vary in their effects on the establishment and growth of microbes introduced to their nectar, as the effects of dispersal by pollinators can mask any effects of microbe filtering by host plants . This hampers our knowledge of whether different plant nectars deterministically select for specific nectar microbes outside of pollinator dispersal, representing a major knowledge gap in our understanding of widespread plant–microbe–pollinator interactions. Empirical plant–pollinator–microbe networks suggest biotic and abiotic filtering mechanisms play a role alongside dispersal in nectar microbe community assembly .

Host plant-related filtering mechanisms may be mediated by floral morphology, as microbial communities can vary even across organs within a single flower . Additionally, chemical constituents of nectar, including sugars, proteins, secondary metabolites and peroxides, have been found to influence microbial growth . Yet, prior studies have not assessed the effects of floral traits per se or plant relatedness on microbial community assembly across diverse plant species and have been limited to a single plant or microbial taxon . Experimentally assessing the assembly of microbe communities across a wide variety of plant species in roots and phyllospheres has yielded significant insights into the processes and traits which shape plant–microbe interactions. Studies including a broader diversity of plant taxa are crucial to better understand the evolution and ecology of how nectar traits mediate plant–microbe interactions. In this study, we leverage a synthetic community of nectarinhabiting bacteria and yeasts, inoculating the floral nectar of 31 plant species. The focal microbes are representative of Northern California flowers and include both nectar specialists and generalists . Our experimental approach provides unique perspectives of nectar microbiome assembly across plant hosts. Using a single uniform community in bagged flowers controls for variation from priority effects and removes pollinator dispersal. Additionally, we control for geography by conducting inoculations in plant species growing within the same general area. By comparing the growth of an initially uniform microbe community across plant species and across seasons due to variation in species’ flowering phenology, we test the following hypotheses: nectar microbes differ in their establishment and growth in nectars of different plant species in the absence of pollinator-mediated dispersal; within a plant species, microbial assembly is deterministic and predicted by floral traits and/or plant phylogenetic relatedness and higher temperatures alter microbial community composition by increasing abundance and favouring growth of certain taxa over others. We found that microbial community assembly in nectar diverged among plant species, in part explained by nectar defence traits, but also with seasonal variation in temperature maxima and minima. Divergence in the trajectory of microbial communities may hold relevance for pollinator visitation and plant reproductive success.We selected five microbe species that are common, widely distributed representatives of nectar microbiomes in various plant species, including those in NorthernCalifornia : the yeasts Metschnikowia reukaufii and Aureobasidium pullulans, and the bacteria Neokomagataea thailandica, Acinetobacter pollinis and Apilactobacillus micheneri. The inoculum contained ~104 cellsμL−1 of each species .We conducted 11 rounds of floral inoculation on the University of California, Davis campus from 22 March to 29 June 2023. In the morning the day before inoculations, we bagged ~10 unopened flower buds on each of 5–8 species of flowering plants to prevent the transfer of microbes by pollinators. We secured organza bags around flowers, removing all open flowers prior to sealing the bag. Each time we handled flowers, we inspected for any breaches by ants or thrips. Flowers that opened within bags were inoculated between 09:00 to 11:00h. To inoculate a flower, 1μL inoculum, carried into the field on ice, was delivered onto the nectary using a micropipette and autoclaved tips , then flowers were tagged with a unique identifier, and re-bagged. Each week, we inoculated roughly five to eight flowers per plant species . Over the course of the study, we recorded temperature extrema for all inoculation days from a local weather station .Roughly 24h after inoculation, we excised flowers from plants, sealed them in containers and transported them to the laboratory. Inside a laminar flow hood, we used glass microcapillary tubes to extract and measure the volume of total nectar in each flower . We quantified microbes in nectar as in Cecala and Vannette . Briefly, we diluted pure nectar in Dulbecco’s phosphate-buffered saline, plated aliquots on each of three agar media types and incubated for 6days, after which CFUs were identified and tallied . Microbial growth from four bagged flowers breached by crawling insects did not differ markedly from that of other flowers and remained in analyses. For each flower, we calculated: the density of CFUs per μL nectar, by dividing the number of CFUs per plate by the actual volume of pure nectar in the aliquot; and the estimated total abundance of CFUs per flower, by multiplying our calculated density by the total volume of nectar originally extracted from that flower.