Automated identification of plants has improved considerably thanks to the recent progress in deep learning and the availability of training data with more and more photos in the field. However, this profusion of data only concerns a few tens of thousands of species, mostly located in North America and Western Europe, much less in the richest regions in terms of biodiversity such as tropical countries. On the other hand, for several centuries, botanists have collected, catalogued and systematically stored plant specimens in herbaria, particularly in tropical regions, and the recent efforts by the biodiversity informatics community made it possible to put millions of digitized sheets online. The LifeCLEF 2020 Plant Identification challenge (or "PlantCLEF 2020") was designed to evaluate to what extent automated identification on the flora of data deficient regions can be improved by the use of herbarium collections. It is based on a dataset of about 1,000 species mainly focused on the South America's Guiana Shield, an area known to have one of the greatest diversity of plants in the world. The challenge was evaluated as a cross-domain classification task where the training set consist of several hundred thousand herbarium sheets and few thousand of photos to enable learning a mapping between the two domains. The test set was exclusively composed of photos in the field. This paper presents the resources and assessments of the conducted evaluation, summarizes the approaches and systems employed by the participating research groups, and provides an analysis of the main outcomes.
Laurinne J Balstad, Joe Brennan, Marissa L. Baskett
et al.
Theoretical ecologists have long leveraged empirical data in various forms to advance ecology. Recently increased volumes and access to ecological data present an expanding set of opportunities for theoreticians to inform model development, framing, and interpretation. Whereas statisticians have collective guidance on best practices for data use, theoreticians might lack formal education on how to integrate diverse types of data into a single ecological model. As a group of predominantly early-career theoretical ecologists, we have developed guiding principles and practical tips to support theoretical ecologists in synthesizing multiple types of data at different phases of the modeling process. Our rules fall into three overarching themes: iteration in the data-model integration process, leveraging multiple sources of data), and understanding uncertainty. Across these rules, we emphasize that the data-model integration requires transparent, justifiable, and defensible communication of modeling choices to support readers in appropriately contextualizing the model and its implications.
Automated plant identification has improved considerably thanks to recent advances in deep learning and the availability of training data with more and more field photos. However, this profusion of data concerns only a few tens of thousands of species, mainly located in North America and Western Europe, much less in the richest regions in terms of biodiversity such as tropical countries. On the other hand, for several centuries, botanists have systematically collected, catalogued and stored plant specimens in herbaria, especially in tropical regions, and recent efforts by the biodiversity informatics community have made it possible to put millions of digitised records online. The LifeCLEF 2021 plant identification challenge (or "PlantCLEF 2021") was designed to assess the extent to which automated identification of flora in data-poor regions can be improved by using herbarium collections. It is based on a dataset of about 1,000 species mainly focused on the Guiana Shield of South America, a region known to have one of the highest plant diversities in the world. The challenge was evaluated as a cross-domain classification task where the training set consisted of several hundred thousand herbarium sheets and a few thousand photos to allow learning a correspondence between the two domains. In addition to the usual metadata (location, date, author, taxonomy), the training data also includes the values of 5 morphological and functional traits for each species. The test set consisted exclusively of photos taken in the field. This article presents the resources and evaluations of the assessment carried out, summarises the approaches and systems used by the participating research groups and provides an analysis of the main results.
Medicinal plants have been a key component in producing traditional and modern medicines, especially in the field of Ayurveda, an ancient Indian medical system. Producing these medicines and collecting and extracting the right plant is a crucial step due to the visually similar nature of some plants. The extraction of these plants from nonmedicinal plants requires human expert intervention. To solve the issue of accurate plant identification and reduce the need for a human expert in the collection process; employing computer vision methods will be efficient and beneficial. In this paper, we have proposed a model that solves such issues. The proposed model is a custom convolutional neural network (CNN) architecture with 6 convolution layers, max-pooling layers, and dense layers. The model was tested on three different datasets named Indian Medicinal Leaves Image Dataset,MED117 Medicinal Plant Leaf Dataset, and the self-curated dataset by the authors. The proposed model achieved respective accuracies of 99.5%, 98.4%, and 99.7% using various optimizers including Adam, RMSprop, and SGD with momentum.
Hanna Herasimchyk, Robin Labryga, Tomislav Prusina
We present a multi-head vision transformer approach for multi-label plant species prediction in vegetation plot images, addressing the PlantCLEF 2025 challenge. The task involves training models on single-species plant images while testing on multi-species quadrat images, creating a drastic domain shift. Our methodology leverages a pre-trained DINOv2 Vision Transformer Base (ViT-B/14) backbone with multiple classification heads for species, genus, and family prediction, utilizing taxonomic hierarchies. Key contributions include multi-scale tiling to capture plants at different scales, dynamic threshold optimization based on mean prediction length, and ensemble strategies through bagging and Hydra model architectures. The approach incorporates various inference techniques including image cropping to remove non-plant artifacts, top-n filtering for prediction constraints, and logit thresholding strategies. Experiments were conducted on approximately 1.4 million training images covering 7,806 plant species. Results demonstrate strong performance, making our submission 3rd best on the private leaderboard. Our code is available at https://github.com/geranium12/plant-clef-2025/tree/v1.0.0.
Plants exchange information through multiple modalities, including chemical, electrical, mycorrhizal, and acoustic signaling, which collectively support survival, defense, and adaptation. While these processes are well documented in biology, their systematic analysis from an Information and Communication Technology (ICT) perspective remains limited. To address this gap, this article is presented as a tutorial with survey elements. It provides the necessary biological background, reformulates inter-plant signaling within ICT frameworks, and surveys empirical studies to guide future research and applications. First, the paper introduces the fundamental biological processes to establish a foundation for readers in communications and networking. Building on this foundation, existing models of emission, propagation, and reception are synthesized for each modality and reformulated in terms of transmitter, channel, and receiver blocks. To complement theory, empirical studies and state-of-the-art sensing approaches are critically examined. Looking forward, the paper identifies open challenges and outlines future research directions, with particular emphasis on the emerging vision of the Internet of Plants (IoP). This paradigm frames plants as interconnected nodes within ecological and technological networks, offering new opportunities for applications in precision agriculture, ecosystem monitoring, climate resilience, and bio-inspired communication systems. By integrating biological insights with ICT frameworks and projecting toward the IoP, this article provides a comprehensive tutorial on plant communication for the communications research community and establishes a foundation for interdisciplinary advances.
ABSTRACT: Sarcotheca macrophylla Blume (Oxalidaceae), an endemic plant species to Borneo, consists of short-styled (S-morph) and long-styled (L-morph) flowers, referred to as heterostyly. However, the character states related to heterostyly have not been fully disclosed in S. macrophylla. Additionally, S. macrophylla has a reproductive issue due to a lack of seeds in fruits. This research aims to examine the floral characteristics and detail of sexual organs (androecium and gynoecium) in this species and to analyze the consequences of heterostyly on its natural pollination. We collect flowers from 34 individuals and investigate the inaccuracy index and all of floral character states. We used light and stereo microscopes to evaluate the heterostyly syndrome through morpho-anatomical observations of the flowers. Our results reveal that the two floral types (S-morph and L-morph) of S. macrophylla possess reciprocal herkogamy with ancillary pollen polymorphism in palynological characters, differences in pollen number, and pollen viability. These characters suggest that S. macrophylla has distyly type. Furthermore, the reproductive organs of the floral morphs appear to contribute differently to reproductive fitness. The S-morph flowers demonstrate enhanced male function through higher pollen production and viability, while the L-morph flowers exhibit enhanced female function through floral architecture that promotes effective pollen reception.
Upstream open reading frame (uORF)-based genetic engineering has emerged as an excellent strategy for improving agronomic traits of crops. Despite its significant potential, the exploration of CRISPR/Cas-based uORF engineering in many crop species remains unexplored, thereby limiting the application of this approach in genetic innovation of important crops. In this study, we focused on uORF-based genome editing to improve abiotic stresses in Brassica napus. The putative uORFs of BnVTC genes which involved in ascorbic acid (AsA) biosynthesis were selected as potential targets. The AsA contents in leaves, buds, and stems were significantly increased in BnVTC2-uORF-edited mutants. The BnVTC2-uORF-edited mutants exhibited tolerance to environmental stresses, such as low temperature, salinity, and drought. No obvious penalty on yield traits were observed between the BnVTC2-uORF-edited lines and WT. VTC2-uORF sequence was highly conserved across the genus Brassica, coupled with the absence of frameshift mutations in the natural germplasm, which suggested that uORF-targeted gene editing alone could be an effective approach for improving abiotic stresses. This research paves the way for the strategic deployment of CRISPR-based uORF engineering to improve the nutritional profile and abiotic stress resistance of oilseed rape.
The western flower thrips, <i>Frankliniella occidentalis</i>, a worldwide insect pest with its polyphagous feeding behavior and capacity to transmit viruses, follows a diurnal rhythmicity driven by expression of the circadian clock genes. However, it remained unclear how the clock signal triggers the thrips behaviors. This study posed a hypothesis that the clock signal modulates cGMP-dependent protein kinase (<i>PKG</i>) activity to mediate the diurnal behaviors. A <i>PKG</i> gene is encoded in <i>F. occidentalis</i> and exhibits high sequence homologies with those of honeybee and fruit fly. Interestingly, its expression followed a diel pattern with high expression during photophase in larvae and adults of <i>F. occidentalis</i>. It is noteworthy that <i>PKG</i> expression was clearly observed in the midgut during photophase but not in scotophase from our fluorescence in situ hybridization analysis. A prediction of protein–protein interaction suggested its functional association with clock genes. To test this functional link, RNA interference (RNAi) of the PKG gene expression was performed by feeding a gene-specific double-stranded RNA, which led to significant alteration of the two clock genes (<i>Clock</i> and <i>Period</i>) in their expression levels. The RNAi treatment caused adverse effects on early-life development and adult fecundity. To further analyze the role of PKG in affecting diurnal behavior, the adult females were continuously observed for a 24 h period with an automatic digitization device to obtain movement parameters and durations (%) in different micro-areas in the observation arena. Diel difference was observed with speed in RNAi-control females at 0.16 mm/s and 0.08 mm/s, in photo- and scotophase, respectively, whereas diel difference was not observed for the <i>PKG</i>-specific RNAi-treated females, which showed 0.07 mm/s and 0.06 mm/s, respectively. The diel difference was also observed in durations (%) in the control females, more strongly in the intermediate area in the observation arena. Speed and durations in the different micro-areas in mid-scotophase were significantly different from most photophase in the control females, while speed was significantly different mainly during late photophase when comparing effects of control and RNAi treatments in each light phase. Three sequential stages consisting of high activity followed by feeding and visiting of micro-areas were observed for the control females. For RNAi-treated females, the three phases were disturbed with irregular speed and visits to micro-areas. These results suggest that PKG is associated with implementing the diurnal behavior of <i>F. occidentalis</i> by interacting with expressions of the circadian clock genes.
Disordered systems theory provides powerful tools to analyze the generic behaviors of highdimensional systems, such as species-rich ecological communities or neural networks. By assuming randomness in their interactions, universality ensures that many microscopic details are irrelevant to system-wide dynamics; but the choice of a random ensemble still limits the generality of results. We show here, in the context of ecological dynamics, that these analytical tools do not require a specific choice of ensemble, and that solutions can be found based only on a fundamental rotational symmetry in the interactions, encoding the idea that traits can be recombined into new species without altering global features. Dynamical outcomes then depend on the spectrum of the interaction matrix as a free parameter, allowing us to bridge between results found in different models of interactions, and extend beyond them to previously unidentified behaviors. The distinctive feature of ecological models is the possibility of species extinctions, which leads to an increased universality of dynamics as the fraction of extinct species increases. We expect that these findings can inform new developments in theoretical ecology as well as for other families of complex systems.
While deep learning has significantly advanced automatic plant disease detection through image-based classification, improving model explainability remains crucial for reliable disease detection. In this study, we apply the Automated Concept-based Explanation (ACE) method to plant disease classification using the widely adopted InceptionV3 model and the PlantVillage dataset. ACE automatically identifies the visual concepts found in the image data and provides insights about the critical features influencing the model predictions. This approach reveals both effective disease-related patterns and incidental biases, such as those from background or lighting that can compromise model robustness. Through systematic experiments, ACE helped us to identify relevant features and pinpoint areas for targeted model improvement. Our findings demonstrate the potential of ACE to improve the explainability of plant disease classification based on deep learning, which is essential for producing transparent tools for plant disease management in agriculture.
Earth observation data have shown promise in predicting species richness of vascular plants ($α$-diversity), but extending this approach to large spatial scales is challenging because geographically distant regions may exhibit different compositions of plant species ($β$-diversity), resulting in a location-dependent relationship between richness and spectral measurements. In order to handle such geolocation dependency, we propose \textit{Spatioformer}, where a novel geolocation encoder is coupled with the transformer model to encode geolocation context into remote sensing imagery. The Spatioformer model compares favourably to state-of-the-art models in richness predictions on a large-scale ground-truth richness dataset (HAVPlot) that consists of 68,170 in-situ richness samples covering diverse landscapes across Australia. The results demonstrate that geolocational information is advantageous in predicting species richness from satellite observations over large spatial scales. With Spatioformer, plant species richness maps over Australia are compiled from Landsat archive for the years from 2015 to 2023. The richness maps produced in this study reveal the spatiotemporal dynamics of plant species richness in Australia, providing supporting evidence to inform effective planning and policy development for plant diversity conservation. Regions of high richness prediction uncertainties are identified, highlighting the need for future in-situ surveys to be conducted in these areas to enhance the prediction accuracy.
Samhita Marri, Arun N. Sivakumar, Naveen K. Uppalapati
et al.
Tracking plant features is crucial for various agricultural tasks like phenotyping, pruning, or harvesting, but the unstructured, cluttered, and deformable nature of plant environments makes it a challenging task. In this context, the recent advancements in foundational models show promise in addressing this challenge. In our work, we propose PlantTrack where we utilize DINOv2 which provides high-dimensional features, and train a keypoint heatmap predictor network to identify the locations of semantic features such as fruits and leaves which are then used as prompts for point tracking across video frames using TAPIR. We show that with as few as 20 synthetic images for training the keypoint predictor, we achieve zero-shot Sim2Real transfer, enabling effective tracking of plant features in real environments.