1. Automated monitoring of insects and other arthropods is vital for ecological research and conservation; yet current image recognition tools often lack generalizability across diverse imaging conditions and struggle with varying specimen sizes. This limits their practical utility and wide adoption by ecologists. We present flatbug, an open-source Python package developed to provide a robust, generalizable and user-friendly solution for arthropod detection and instance segmentation. 2. flatbug employs an adaptive tiling framework for both training and inference, enabling scale- and size-agnostic detection. It leverages instance segmentation, allowing for automated background removal and precise body-size estimation. The package includes pre-trained models and is accompanied by a uniquely diverse benchmark dataset of over 113,000 annotated arthropods from laboratory- and field-based imaging systems. This dataset facilitates rigorous cross-validation and evaluation. 3. Our best flatbug model achieves an average F1 score of 94.2% across these diverse datasets. Crucially, it demonstrates strong out-of-the-box generalizability, with performance reduced by only 7.1% on average when tested against entirely novel imaging systems excluded from its training. This confirms flatbug's robust performance in new contexts without retraining. 4. flatbug offers ecologists and practitioners a ready-to-use, efficient and accurate tool for arthropod monitoring that addresses common limitations of existing methods. With comprehensive documentation, tutorials and an online demo, it is designed for straightforward integration and use. By providing a generalizable solution and a new standard for evaluating cross-domain performance, flatbug aims to accelerate advancements in automated arthropod detection and ecological computer vision. The package, dataset and models are freely available at https://github.com/darsa-group/flat-bug/ and https://doi.org/10.5281/zenodo.18164125.

Background Arthropods make up the majority of species on Earth. To study their diversity and ecological roles in ecosystems, bulk sampling is commonly used to collect large numbers of specimens. Processing these samples is labor-intensive and time-consuming, often delaying timely decision-making in ecosystem monitoring. Automated detection systems offer a promising alternative to support sample processing; however, most existing systems still have two major limitations. First, the pipelines for localizing and classifying arthropods in images, whether single- or double-stage, are limited. Second, the prediction results often lack information about functional roles. Therefore, we developed InsectRoleVision, a more expert-centered and reliable system that enables inference of arthropod diversity based on their functional roles. Methodology To develop the system, an image dataset with taxonomic resolution was designed and created to support conclusions about the functional roles of the animals. Both single-stage and double-stage recognition pipelines were compared. For single-stage detection and the first stage of double-stage detection, four YOLO models and a transformer were evaluated to localize and classify the arthropods in each image. In double-stage detection, the region of interest (RoI) was cropped into individual images after localization and used to compare four classification models: InceptionV3, ResNet, MobileNet, and VGG19. A logic block pipeline was connected to the prediction results to further infer the richness and proportionality of each class or taxon with respect to their functional roles. Result YOLOv11 was the best-performing model, achieving over 93% mAP, precision, and recall in localizing arthropods in the images. InceptionV3 was the best-performing classifier, achieving 80% precision and recall in classifying more than 43,000 cropped images of arthropods. There was no significant difference between the results of single- and modular double-stage detection strategies. Therefore, the choice between strategies depends on the intended application: single-stage detection provides real-time results and is suitable for real-time detection applications, while double-stage detection allows a human expert to review the detection proposal and refine the classification result. InsectRoleVision has adopted the YOLOv11-InceptionV3 architecture, which is more flexible and human-centered, allowing quick access to both arthropod diversity and ecological roles.

There is growing evidence that human-induced climate change and habitat loss are having negative impacts on insect populations. New technologies have a vital role in improving and expanding global biodiversity monitoring capacity to understand where change is happening and to support restorationMonitoring of insects traditionally needs entomologists in the field, but insect camera traps powered by AI are emerging as a scalable approach to monitoring semi-autonomously. These systems attract, detect, and identify insects using a Raspberry Pi, camera and UV lights. AI algorithms are also being developed by a network of researchers across the world to help with identification, notably in Europe and North America.The first version of a system for monitoring nocturnal insects was developed by Bjerge et al. 2021. An Automated Light Trap to Monitor Moths (Lepidoptera) Using Computer Vision-Based Tracking and Deep Learning. Here we describe the second generation of the system as an open-source solution. This paper aims to enable anyone to build their own system, and to iterate and improve the design for their needs. This system captures images at set intervals or based on motion detection to monitor insects that are attracted to lights at night. The UKCEH Automated Monitoring of Insects System (UKCEH AMI-system) is an insect camera trap designed using a single board computer, USB camera and attractant lights as the primary components along with peripheral accessories to make an autonomous system capable of long-term deployment in the field. Nearly 200 UKCEH AMI-systems have been deployed to date in over 30 countries around the world.

Abstract Automated monitoring of insect pollinators in natural environments with insect camera traps and trained deep learning algorithms provides novel data for insect ecological studies. However, efficient and accurate image recognition analysis of the recorded images or videos is challenging, particularly for images containing small insects against complex backgrounds with diverse vegetation communities. Even when insects can be detected in images, identifying their taxonomy remains difficult, particularly in footage with low image resolution, light conditions, and distances from the plants, and in cases where insects appear blurry or only partially visible. In this work, we present InsectDCT , an AI-based pipeline for automated detection, hierarchical classification, and tracking of insects in footage of natural vegetation tested in different environments. The InsectDCT pipeline consists of three levels: insect Detection and localization, hierarchical taxonomic Classification, and spatio-temporal Tracking. In the first stage, insects are detected in time-lapse images or video recordings using the You Only Look Once (YOLO11) object detection architecture. Detection performance is improved using motion-enhanced images, which improve robustness in cluttered and 3 dimensional environments. The detector is trained on an extensive dataset that contains more than 60,000 images collected using camera traps deployed across a wide range of plant families and floral habitats. In the second stage, detected insects are classified using a hierarchical taxonomy-aware classification framework that covers 80 taxonomic groups. Classification is performed at multiple taxonomic levels, including order, family, and genus/species, allowing coarse and fine-grained ecological analyzes while accounting for varying levels of visual ambiguity. In the third stage, a multi-object tracking module is applied to high temporal-resolution image sequences and video data to associate detections of the same individual across time. InsectDCT code and all datasets are made publicly available.

In recent years, insects have gained significant attention as a potential alternative nutritional source to address global food security and sustainability challenges. Commercial insect production is thus gaining traction and various insect species are being farmed for food and feed. Despite the increasing interest in commercial insect production and its promising future, there is still much room for improvement in terms of production efficiency. The profitability of insect production for food and feed relies on products’ nutritional content, composition, and safety, which varies greatly according to insect species, diet, and development stage during harvest. Consequently, maintaining consistent nutritional composition of insect products require rapid and cost-effective methods for the determination of nutritional content and safety of insects. Additionally, a promising strategy for enhancing the nutritional quality of insect products may involve implementing selective breeding. This would necessitate employing non-destructive methods to measure the protein and fat content on all live selection candidates. This review paper presents a comprehensive study of the reference methods employed in determining the nutritional composition of edible insects. While highly accurate, the gold standard chemometric methods are widely considered to be costly, time-consuming, and destructive for determining the nutrition content. Our review paper therefore further extends to emerging technologies leveraging spectroscopy, computer vision, and spectral imaging tools that promise fast and cost-effective prediction of the nutritional composition and safety of edible insects and insect products. The review paper also evaluates the suitability of these tools for predicting nutritional composition and safety of insect products providing examples of their applications in various contexts within the insect-based food and feed industry.

The growing interest in insect farming as a sustainable protein alternative has given rise to the commercial production of key species like the Black Soldier Fly (BSF), primarily for use in livestock, fish, and pet nutrition. Despite the heightened interest in BSF production, there is a need for increased efficiency, particularly in the context of large-scale measurement of various traits for selective breeding as well as management optimization. The unique insect production systems, coupled with the challenges posed by their small size, fragility, and metamorphic life cycle underscores the necessity for innovative approaches to streamline production.This study explores the potential of computer vision (CV) in predicting the larval sex and morphological traits of BSF, offering a non-invasive, rapid, and automated method for trait measurement. The study explores algorithms utilizing You-Only-Look-Once (YOLOv8) in detection and segmentation, ResNet for feature extraction and classification, and regression analysis mechanisms. We assess the ability of our models to predict larval weight from images through morphometric weight prediction and CNN-regression approaches.A notable contribution of this study is the pioneering effort to classify BSF larval sex using CV and deep learning (DL). In the analysis of larval weight prediction, a coefficient determination (R2) of up to 0.80 between measured and predicted weight was achieved using the morphometric weight prediction approach, along with an R2 of 0.71 through the CNN-regression approach. Additionally, the sex prediction module demonstrated an F1 score of 0.75 and a prediction accuracy of 74 %. These results underscore the feasibility of leveraging CV techniques for predicting the sex and body traits of BSF larvae, representing a significant advancement toward the automation of selective breeding in the context of insect farming.

Computer vision methods offer great potential for rapid image-based identification of medically important arthropod specimens. However, imaging large numbers of specimens is time consuming, and it is difficult to achieve the high image quality required for machine learning models. Conventional imaging methods for identifying and digitizing arthropods, such as insects and spiders, use a stereomicroscope or macro lenses with a camera. This method is challenging due to the narrow field of view, especially when large numbers of arthropods need to be processed. In this paper, we present a high-throughput scanner-based method for capturing images of arthropods that can be used to generate large datasets suitable for training machine learning algorithms for identification. We demonstrate the ability of this approach to image arthropod samples collected with different sampling methods, such as sticky traps (unbaited, in different colors), baited mosquito traps as used by the US Centers for Disease Control and Prevention (CDC) and BioGents-Sentinel (BGS), and UV light traps with a sticky pad. Using different strategies to place the arthropods on a charge-coupled device (CCD) flatbed scanner and optimized settings that balance processing time and image quality, we captured high-resolution images of various arthropods and obtained morphological details with resolution and magnification similar to a stereomicroscope. We validate the method by comparing the performance of three different deep learning models (InceptionV3, ResNet and MobileNetV2) on two different datasets, namely the scanned images from this study and the images captured with a camera of a stereomicroscope. The results show that the performance of the models trained on the two datasets is not significantly different, indicating that the quality of the scanned images is comparable to that of a stereomicroscope.

Insects represent nearly half of all known multicellular species, but knowledge about them lags behind for most vertebrate species. In part for this reason, they are often neglected in biodiversity conservation policies and practice. Computer vision tools, such as insect camera traps, for automated monitoring have the potential to revolutionize insect study and conservation. To further advance insect camera trapping and the analysis of their image data, effective image processing pipelines are needed. In this paper, we present a flexible and fast processing pipeline designed to analyse these recordings by detecting, tracking and classifying nocturnal insects in a broad taxonomy of 15 insect classes and resolution of individual moth species. A classifier with anomaly detection is proposed to filter dark, blurred or partially visible insects that will be uncertain to classify correctly. A simple track-by-detection algorithm is proposed to track classified insects by incorporating feature embeddings, distance and area cost. We evaluated the computational speed and power performance of different edge computing devices (Raspberry Pi's and NVIDIA Jetson Nano) and compared various time-lapse (TL) strategies with tracking. The minimum difference of detections was found for 2-min TL intervals compared to tracking with 0.5 frames per second; however, for insects with fewer than one detection per night, the Pearson correlation decreases. Shifting from tracking to TL monitoring would reduce the number of recorded images and would allow for edge processing of images in real-time on a camera trap with Raspberry Pi. The Jetson Nano is the most energy-efficient solution, capable of real-time tracking at nearly 0.5 fps. Our processing pipeline was applied to more than 5.7 million images recorded at 0.5 frames per second from 12 light camera traps during two full seasons located in diverse habitats, including bogs, heaths and forests. Our results thus show the scalability of insect camera traps.

Deep clustering has proven successful in analyzing complex, high-dimensional real-world data. Typically, features are extracted from a deep neural network and then clustered. However, training the network to extract features that can be clustered efficiently in a semantically meaningful way is particularly challenging when data is sparse. In this paper, we present a semi-supervised method to fine-tune a deep learning network using Model-Agnostic Meta-Learning, commonly employed in Few-Shot Learning. We apply episodic training with a novel multivariate scatter loss, designed to enhance inter-class feature separation while minimizing intra-class variance, thereby improving overall clustering performance. Our approach works with state-of-the-art deep learning models, spanning convolutional neural networks and vision transformers, as well as different clustering algorithms like K-means and Spectral clustering. The effectiveness of our method is tested on several commonly used Few-Shot Learning datasets, where episodic fine-tuning with our multivariate scatter loss and a ConvNeXt backbone outperforms other models, achieving adjusted rand index scores of 89.7% on the EU moths dataset and 86.9% on the Caltech birds dataset, respectively. Hence, our proposed method can be applied across various practical domains, such as clustering images of animal species in biology.

To better understand the status and trends of insects and other arthropods, emerging technologies like image recognition are developing rapidly. This is creating a strong demand for efficient and accurate algorithms for detection and localization of arthropods in images. Existing models have modest performance and do not generalise well to variation in scale, appearance and density of specimens, or imaging conditions. Consequently, each new application often requires manual labeling of training data and model training, which limits the uptake of image-based tools and technologies. Here, we introduce flatbug, which is a powerful and general model to count and outline insects and other terrestrial arthropods in images. The training dataset is large and diverse and represent 23 different lab- and field-based imaging systems. The best flatbug model achieves an average F 1 = 94.2% on our validation dataset. Crucially, we show that flatbug has great out-of-the-box performance and generalises well to novel contexts. When images from a given dataset are left out of model training, the performance of flatbug is only reduced by on average 7.1% for the dataset in question. By using truly stratified cross-validation, we set a precedent for robust evaluation of deep learning model performance and generalization. We also take steps towards scale- and size-agnostic arthropod detection, by developing an integrated tiling framework for inference and training. Additionally, flatbug ‘s implementation of YOLOv8 for instance segmentation enables downstream background removal and body size estimation. The generaliseability of flatbug stems from the diversity of contexts represented in the flatbug dataset, including 113550 arthropods annotated across 6131 images. Alongside a fully documented Python package with tutorials for integration and analysis via , the flatbug dataset is available from . By providing performant models and the accompanying dataset, flatbug offers both a ready-to-use tool and a benchmark for the future. Overall, flatbug represents a significant methodological advance within arthropod image detection, with user-friendly integration for monitoring and research. ![Figure][1] ### Competing Interest Statement The authors have declared no competing interest. [1]: pending:yes

Addressing global declines in insect biodiversity requires both ecological restoration and high-quality monitoring data. While long-term participatory schemes have been foundational, recent advances in automated recording and AI-based identification offer transformative but undocumented potential. Here, we show how leveraging insect camera traps, deep learning models and statistics drives a step-change in ecological knowledge. We highlight four key areas of ecological understanding: phenology, abundance, richness, and community dynamics, and show how automated data can correct phenological estimates by weeks and improve biodiversity assessments. Data from insect camera traps offer unprecedented resolution and scalability, making them powerful tools for tracking insect communities and informing conservation strategies.

Arthropods, including insects, represent the most diverse group and contribute significantly to animal biomass. Automatic monitoring of insects and other arthropods enables quick and efficient observation and management of ecologically and economically important targets such as pollinators, natural enemies, disease vectors, and agricultural pests. The integration of cameras and computer vision facilitates innovative monitoring approaches for agriculture, ecology, entomology, evolution, and biodiversity. However, studying insects and their interactions with flowers and vegetation in natural environments remains challenging, even with automated camera monitoring.This paper presents a comprehensive methodology to monitor abundance and diversity of arthropods in the wild and to quantify floral cover as a key resource. We apply the methods across more than 10 million images recorded over two years using 48 insect camera traps placed in three main habitat types. The cameras monitor arthropods, including insect visits, on a specific mix of Sedum plant species with white, yellow and red/pink colored of flowers. The proposed deep-learning pipeline estimates flower cover and detects and classifies arthropod taxa from time-lapse recordings. However, the flower cover serves only as an estimate to correlate insect activity with the flowering plants.Color and semantic segmentation with DeepLabv3 are combined to estimate the percent cover of flowers of different colors. Arthropod detection incorporates motion-informed enhanced images and object detection with You-Only-Look-Once (YOLO), followed by filtering stationary objects to minimize double counting of non-moving animals and erroneous background detections. This filtering approach has been demonstrated to significantly decrease the incidence of false positives, since arthropods, occur in less than 3% of the captured images.The final step involves grouping arthropods into 19 taxonomic classes. Seven state-of-the-art models were trained and validated, achieving F1-scores ranging from 0.81 to 0.89 in classification of arthropods. Among these, the final selected model, EfficientNetB4, achieved an 80% average precision on randomly selected samples when applied to the complete pipeline, which includes detection, filtering, and classification of arthropod images collected in 2021. As expected during the beginning and end of the season, reduced flower cover correlates with a noticeable drop in arthropod detections. The proposed method offers a cost-effective approach to monitoring diverse arthropod taxa and flower cover in natural environments using time-lapse camera recordings.

Insects represent half of all global biodiversity, yet many of the world's insects are disappearing, with severe implications for ecosystems and agriculture. Despite this crisis, data on insect diversity and abundance remain woefully inadequate, due to the scarcity of human experts and the lack of scalable tools for monitoring. Ecologists have started to adopt camera traps to record and study insects, and have proposed computer vision algorithms as an answer for scalable data processing. However, insect monitoring in the wild poses unique challenges that have not yet been addressed within computer vision, including the combination of long-tailed data, extremely similar classes, and significant distribution shifts. We provide the first large-scale machine learning benchmarks for fine-grained insect recognition, designed to match real-world tasks faced by ecologists. Our contributions include a curated dataset of images from citizen science platforms and museums, and an expert-annotated dataset drawn from automated camera traps across multiple continents, designed to test out-of-distribution generalization under field conditions. We train and evaluate a variety of baseline algorithms and introduce a combination of data augmentation techniques that enhance generalization across geographies and hardware setups. Code and datasets will be made publicly available.

Insects represent nearly half of all known organisms, with nocturnal insects being particularly challenging to monitor. Computer vision tools for automated monitoring have the potential to revolutionize insect study and conservation. The advancement of light traps with camera-based monitoring systems for insects necessitates effective and flexible pipelines for analysing recorded images. In this paper, we present a flexible and fast processing pipeline designed to analyse these recordings by detecting, tracking and classifying nocturnal insects at the taxonomic ranks of order, suborder as well as the resolution of individual moth species. The pipeline consists of four adaptable steps. The first step detect insects in the camera trap images. An order and suborder classifier with anomaly detection is proposed to filter dark, blurry or partly visible insects that will be uncertain to classify correctly. A simple track-by-detection algorithm is proposed to track the classified insects by incorporating feature embeddings, distance and area cost. We evaluated the computational speed and power performance of different edge computing devices (Raspberry Pi's and NVIDIA Jetson Nano) and compared various time-lapse strategies with tracking. The minimum difference was found for 2-minute time-lapse intervals compared to tracking with 0.5 frames per second, however, for insects with fewer than one detection per night, the Pearson correlation decreases. Shifting from tracking to time-lapse monitoring would reduce the amount of recorded images and be able to perform edge processing of images in real-time on a camera trap with Raspberry Pi. The Jetson Nano is the most energy-efficient solution, capable of real-time tracking at nearly 0.5 fps. Our processing pipeline was applied to more than 3.4 million images recorded at 0.5 frames per second from 12 light camera traps during one full season located in diverse habitats, including bogs, heaths and forests. ### Competing Interest Statement The authors have declared no competing interest.

Commercial insect production is a relatively new field that has gained traction in recent years due to its potential as a sustainable source of protein. Despite its promising future, the industry is still in its infancy, and there is much room for improvement in terms of production efficiency. To achieve this, it is essential to implement advanced technologies that can aid in process management. Recent progress in fields such as computer vision (CV) and machine learning has opened up numerous possibilities within insect rearing, encompassing automatic detection, identification, classification, as well as monitoring and tracking. These applications find relevance in automating insect production processes, ensuring insect product quality as well as environmental monitoring and control. The primary objective of this article is to highlight the potential of CV and deep learning (DL) in the domain of insect production for food and feed. It provides an in-depth overview of the key developments in this domain, shedding light on both challenges and opportunities. The article also presents various systems, accompanied by real-world examples and recent advancements, including the integration of machine learning. In conclusion, the article underscores the substantial potential of CV and machine learning to enhance the efficiency and productivity of insect production while identifying areas that warrant further research to advance the insect production sector.

Machine learning has achieved considerable success in data-intensive applications, yet encounters challenges when confronted with small datasets. Recently, few-shot learning (FSL) has emerged as a promising solution to address this limitation. By leveraging prior knowledge, FSL exhibits the ability to swiftly generalize to new tasks, even when presented with only a handful of samples in an accompanied support set. This paper extends the scope of few-shot learning by incorporating novelty detection for samples of categories not present in the support set of FSL. This extension holds substantial promise for real-life applications where the availability of samples for each class is either sparse or absent. Our approach involves adapting existing FSL methods with a cosine similarity function, complemented by the learning of a probabilistic threshold to distinguish between known and outlier classes. During episodic training with domain generalization, we introduce a scatter loss function designed to disentangle the distribution of similarities between known and outlier classes, thereby enhancing the separation of novel and known classes. The efficacy of the proposed method is evaluated on commonly used FSL datasets and the EU Moths dataset characterized by few samples. Our experimental results showcase accuracy, ranging from 95.4

Cameras and computer vision are revolutionising the study of insects, creating new research opportunities within agriculture, epidemiology, evolution, ecology and monitoring of biodiversity. However, the diversity of insects and close resemblances of many species are a major challenge for image-based species-level classification. Here, we present an algorithm to hierarchically classify insects from images, leveraging a simple taxonomy to (1) classify specimens across multiple taxonomic ranks simultaneously, and (2) identify the lowest rank at which a reliable classification can be reached. Specifically, we propose multitask learning, a loss function incorporating class dependency at each taxonomic rank, and anomaly detection based on outlier analysis to quantify the uncertainty. First, we compile a dataset of 41,731 images of insects, combining images from time-lapse monitoring of floral scenes with images from the Global Biodiversity Information Facility (GBIF). Second, we adapt state-ofthe-art convolutional neural networks, ResNet and EfficientNet, for the hierarchical classification of insects belonging to three orders, five families and nine species. Third, we assess model generalization for 11 species unseen by the trained models. Here, anomaly detection is used to predict the higher rank of the species which were not present in the training set. We found that incorporating a simple taxonomy into our model increased the accuracy at higher taxonomic ranks. As expected, our algorithm correctly classified new insect species at higher taxonomic ranks, while classification was uncertain at lower taxonomic ranks. Anomaly detection can effectively flag novel taxa that are visually distinct from species in the training data. However, five novel taxa were consistently mistaken for visually similar species in the training data. Above all, we have demonstrated a practical approach to hierarchical classification based on species taxonomy and uncertainty during automated in situ monitoring of live insects. Our method is simple and versatile, forming a valuable step towards high-level classification of species not found in training data.

As pollinators, insects play a crucial role in ecosystem management and world food production. However, insect populations are declining, necessitating efficient insect monitoring methods. Existing methods analyze video or time-lapse images of insects in nature, but analysis is challenging as insects are small objects in complex and dynamic natural vegetation scenes. In this work, we provide a dataset of primarily honeybees visiting three different plant species during two months of the summer. The dataset consists of 107,387 annotated time-lapse images from multiple cameras, including 9423 annotated insects. We present a method for detecting insects in time-lapse RGB images, which consists of a two-step process. Firstly, the time-lapse RGB images are preprocessed to enhance insects in the images. This motion-informed enhancement technique uses motion and colors to enhance insects in images. Secondly, the enhanced images are subsequently fed into a convolutional neural network (CNN) object detector. The method improves on the deep learning object detectors You Only Look Once (YOLO) and faster region-based CNN (Faster R-CNN). Using motion-informed enhancement, the YOLO detector improves the average micro F1-score from 0.49 to 0.71, and the Faster R-CNN detector improves the average micro F1-score from 0.32 to 0.56. Our dataset and proposed method provide a step forward for automating the time-lapse camera monitoring of flying insects.

High-resolution monitoring is fundamental to understand ecosystems dynamics in an era of global change and biodiversity declines. While real-time and automated monitoring of abiotic components has been possible for some time, monitoring biotic components-for example, individual behaviours and traits, and species abundance and distribution-is far more challenging. Recent technological advancements offer potential solutions to achieve this through: (i) increasingly affordable high-throughput recording hardware, which can collect rich multidimensional data, and (ii) increasingly accessible artificial intelligence approaches, which can extract ecological knowledge from large datasets. However, automating the monitoring of facets of ecological communities via such technologies has primarily been achieved at low spatiotemporal resolutions within limited steps of the monitoring workflow. Here, we review existing technologies for data recording and processing that enable automated monitoring of ecological communities. We then present novel frameworks that combine such technologies, forming fully automated pipelines to detect, track, classify and count multiple species, and record behavioural and morphological traits, at resolutions which have previously been impossible to achieve. Based on these rapidly developing technologies, we illustrate a solution to one of the greatest challenges in ecology: the ability to rapidly generate high-resolution, multidimensional and standardised data across complex ecologies.

Image-based methods for species identification offer cost-efficient solutions for biomonitoring. This is particularly relevant for invertebrate studies, where bulk samples often represent insurmountable workloads for sorting, identifying, and counting individual specimens. On the other hand, image-based classification using deep learning tools have strict requirements for the amount of training data, which is often a limiting factor. Here, we examine how classification accuracy increases with the amount of training data using the BIODISCOVER imaging system constructed for image-based classification and biomass estimation of invertebrate specimens. We use a balanced dataset of 60 specimens of each of 16 taxa of freshwater macroinvertebrates to systematically quantify how classification performance of a convolutional neural network (CNN) increases for individual taxa and the overall community as the number of specimens used for training is increased. We show a striking 99.2% classification accuracy when the CNN (EfficientNet-B6) is trained on 50 specimens of each taxon, and also how the lower classification accuracy of models trained on less data is particularly evident for morphologically similar species placed within the same taxonomic order. Even with as little as 15 specimens used for training, classification accuracy reached 97%. Our results add to a recent body of literature showing the huge potential of image-based methods and deep learning for specimen-based research, and furthermore offers a perspective to future automatized approaches for deriving ecological data from bulk arthropod samples.