Figure 3 - uploaded by Angela Martin
Content may be subject to copyright.
Evaluation metrics accuracy, precision, recall, F-score, and Intersection over Union (IoU) for classifications, predictions, object detections, and segmentations.

Evaluation metrics accuracy, precision, recall, F-score, and Intersection over Union (IoU) for classifications, predictions, object detections, and segmentations.

Source publication
Article
Full-text available
The deep learning (DL) revolution is touching all scientific disciplines and corners of our lives as a means of harnessing the power of big data. Marine ecology is no exception. New methods provide analysis of data from sensors, cameras, and acoustic recorders, even in real time, in ways that are reproducible and rapid. Off-the-shelf algorithms fin...

Context in source publication

Context 1
... evaluate the performance of a trained model, different parameters are utilized by the different approaches, such as accuracy, precision, and recall ( Figure 3). Accuracy is the ratio of correct classifications to the total number of classifications. ...

Citations

... image analysis, metabarcoding) to track changes in the composition and abundance of the entire planktonic community is now the goal of many monitoring programs embracing ecosystembased management (Lombard et al. 2019, Garcia-Vazquez et al. 2021. Automatic processing of plankton samples using image analysis technologies and machine-learning algorithms is one of the costefficient alternatives to traditional time-consuming and costly microscopy (Goodwin et al. 2022, Irisson et al. 2022. These methods allow the processing of a larger number of samples, albeit at a lower taxonomic resolution than microscopy currently allows (i.e. ...
Article
Full-text available
Plankton dynamics in temperate ecosystems have been mainly studied during productive seasons, with comparatively less research conducted during the winter, particularly on microplankton. Implementing plankton sampling during a regular fishery cruise, we investigated the North Sea micro- and mesozooplankton community composition, abundance and size structure (55-2000 µm) during autumn (Buchan/Banks area) and winter (Downs area) between 2013 and 2019. Samples were analyzed using image-based techniques. Community diversity (broad taxa) was relatively similar across years in both areas, with diatoms and tripos taxa sets dominating the microplankton community and gastropods and copepods the mesozooplankton one. The average micro- to mesoozooplankton ratio (in abundance) was 90:1 for Buchan/Banks, resulting in average Normalized Abundance Size Spectra (NASS) slopes of -1.45 ±0.18 SD. For Downs, the micro- to mesoozooplankton ratio was 235:1 and steeper NASS slopes of -1.67 ±0.20 SD due to a lower contribution of large organisms. Interannual changes in the planktonic community for each area and their potential environmental drivers were examined using a redundancy analysis (including taxonomy and size) and a correlation analysis using NASS slopes (size only). Both approaches highlighted the importance of water mass properties (e.g. salinity, temperature, turbidity) in shaping plankton dynamics, although the amount of explained variance differed between approaches (11 versus 46%). Our results contribute to a better understanding of standing stocks of plankton and their environmental drivers. Specifically, novel insights were gained into microplankton dynamics, which play an important role in supporting the growth and survival of winter-spawned fish larvae in the North Sea.
... The F1-score of 1 reflects a perfect balance between precision and recall. Higher values generally indicate better model performance, with 1 being the ideal [61]. These metrics offer a thorough evaluation of the capacity of the model to discern health conditions in robotic systems. ...
Article
Full-text available
Due to their precision, compact size, and high torque transfer, Rotate vector (RV) reducers are becoming more popular in industrial robots. However, repetitive operations and varying speed conditions mean that these components are prone to mechanical failure. Therefore, it is important to develop effective health monitoring (HM) strategies. Traditional approaches for HM, including those using vibration and acoustic emission sensors, encounter such challenges as noise interference, data inconsistency, and high computational costs. Deep learning-based techniques, which use current electrical data embedded within industrial robots, address these issues, offering a more efficient solution. This research provides transfer learning (TL) models for the HM of RV reducers, which eliminate the need to train models from scratch. Fine-tuning pre-trained architectures on operational data for the three different reducers of health conditions, which are healthy, faulty, and faulty aged, improves fault classification across different motion profiles and variable speed conditions. Four TL models, EfficientNet, MobileNet, GoogleNet, and ResNET50v2, are considered. The classification accuracy and generalization capabilities of the suggested models were assessed across diverse circumstances, including low speed, high speed, and speed fluctuations. Compared to the other models, the proposed EfficientNet model showed the most promising results, achieving a testing accuracy and an F1-score of 98.33% each, which makes it best suited for the HM of robotic reducers.
... With improving possibilities of digital image acquisition and analysis, methods combining medium-to large-scale image data collection with deep neural networks have recently spread rapidly in biodiv ersity r esearc h [25][26][27], including in the aquatic and micr oscopic r ealm [ 28 , 29 ]. In the case of diatoms, although not yet br oadl y a pplied, slide scanning micr oscop y no w provides a possibility of large-scale digital image acquisition suitable for the standard type of diatom pr epar ations [ 18 , 30-34 ]. ...
Article
*** OPEN ACCESS ARTICLE *** Diatoms are microalgae with finely ornamented microscopic silica shells. Their taxonomic identification by light microscopy is routinely used as part of community ecological research as well as ecological status assessment of aquatic ecosystems, and a need for digitalization of these methods has long been recognized. Alongside their high taxonomic and morphological diversity, several other factors make diatoms highly challenging for deep learning–based identification using light microscopy images. These include (i) an unusually high intraclass variability combined with small between-class differences, (ii) a rather different visual appearance of specimens depending on their orientation on the microscope slide, and (iii) the limited availability of diatom experts for accurate taxonomic annotation. Findings We present the largest diatom image dataset thus far, aimed at facilitating the application and benchmarking of innovative deep learning methods to the diatom identification problem on realistic research data, “UDE DIATOMS in the Wild 2024.” The dataset contains 83,570 images of 611 diatom taxa, 101 of which are represented by at least 100 examples and 144 by at least 50 examples each. We showcase this dataset in 2 innovative analyses that address individual aspects of the above challenges using subclustering to deal with visually heterogeneous classes, out-of-distribution sample detection, and semi-supervised learning. Conclusions The problem of image-based identification of diatoms is both important for environmental research and challenging from the machine learning perspective. By making available the so far largest image dataset, accompanied by innovative analyses, this contribution will facilitate addressing these points by the scientific community.
... The computing complexity as detailed in Table 1 of each approach varies considerably depending on the environment and the quantity of integrated sensor modalities. The following table delineates the trade-offs among various methods: Underwater environments that are complex and data-rich [126][127][128] b. Quantitative evaluation of sensor fusion techniques ...
Article
Full-text available
Underwater simultaneous localization and mapping (SLAM) has significant challenges due to the complexities of underwater environments, marked by limited visibility, variable conditions, and restricted global positioning system (GPS) availability. This study provides a comprehensive analysis of sensor fusion techniques in underwater SLAM, highlighting the amalgamation of proprioceptive and exteroceptive sensors to improve UUV navigational accuracy and system resilience. Essential sensor applications, including inertial measurement units (IMUs), Doppler velocity logs (DVLs), cameras, sonar, and LiDAR (light detection and ranging), are examined for their contributions to navigation and perception. Fusion methodologies, such as Kalman filters, particle filters, and graph-based SLAM, are evaluated for their benefits, limitations, and computational demands. Additionally, innovative technologies like quantum sensors and AI-driven filtering techniques are examined for their potential to enhance SLAM precision and adaptability. Case studies demonstrate practical applications, analyzing the compromises between accuracy, computational requirements, and adaptability to environmental changes. This paper proceeds to emphasize future directions, stressing the need for advanced filtering and machine learning to address sensor drift, noise, and environmental unpredictability, hence improving autonomous underwater navigation through reliable sensor fusion.
... The scientific community relies on updated information about species distribution and accurate data on ecosystems to develop effective management plans and conservation strategies (Goodwin et al., 2022). A promising approach to meet this need is the integration of citizen science with advanced image recognition and classification tools. ...
... This approach results in more comprehensive datasets, containing images, videos, audio, and information on the spatial distribution of species. Furthermore, it allows for improving classification accuracy and reducing error rates in models through the incorporation of these new data generated (Lopez-Vazquez et al., 2020;Goodwin et al., 2022). Ultimately, this integration of citizen science and imaging technology not only strengthens the scientific knowledge base, but also empowers the scientific community to develop more efficient management plans and conservation strategies, based on comprehensive and accurate data. ...
... In addition to the increasing role of citizen science, there have been significant advancements in observation methods driven by improvements in technology that have led to the development of various tools. Moreover, several research fields are experiencing rapid transformations thanks to the utilization of Artificial Intelligence (AI) for data interpretation (Goodwin et al., 2022). Among these interpretation techniques, "machine learning" (ML) has garnered considerable attention in the scientific community. ...
Article
The genus Arctocephalus represents the group of fur seals that mainly inhabit the Southern Hemisphere. In general, Arctocephalus species are extremely similar in appearance, often making it very difficult to impossible to distinguish them only by characteristics of their external morphology. In this context, it is important to find new tools to differentiate them, especially in locations outside of their traditional distribution area, such as Brazilian waters, in order to take appropriate actions for their management. This study proposes the use of an artificial intelligence method, based on machine learning and convolutional neural networks, to classify and identify three species of southern fur seals by analysing 121 facial images from living specimens of Arctocephalus australis, A. gazella, and A. tropicalis found on the Brazilian coast. The image database Keywords: citizen science convolutional neural networks, machine learning, pinnipeds.
... The prediction of various seawater quality parameters can assist in the early warning of harmful algal blooms [49], and the correlation between parameters can help to analyze the interactions between parameters. The above work provides important data support and decision-making references for the subsequent marine research [50]. ...
Article
Full-text available
In recent years, with the increasing pollution of near-shore waters, the water quality pollution incidents have been aggravated, which seriously threatens many aspects of coastal economic development, ecological environment and living health. Therefore, there is an urgent need for an effective method to predict the water quality of near-shore waters. However, due to seasonal changes, ocean currents, biological activities and other factors, the marine environment has strong complexity and uncertainty, which leads to the monitoring data of seawater quality parameters are unstable, non-linear and other characteristics. At the same time, there are interactions between different parameters, so it is not easy to dig deeper into the information in the data, and the accuracy of the existing prediction methods for multi-parameter multi-step prediction of seawater quality is generally low. To solve the above problems, a new graph neural network model is proposed in this paper. The model can effectively extract the local time correlation, global time correlation and spatial correlation in non-Euclidean space of seawater quality parameter data from multiple dimensions. Finally, this paper evaluates the model performance using the seawater parameter data from the near-shore waters of Beibu Gulf, and compared with the five baseline models, the model proposed in this paper shows the best performance in all the defined evaluation indexes.
... More recently, the use of artificial intelligence (machine learning, and especially deep learning), has demonstrated its efficiency across various domains, including image and speech processing (LeCun et al., 2015;Shinde and Shah, 2018). With the significant breakthrough of deep learning, the automation of monitoring processes in bioacoustics has become a realistic objective (Goodwin et al., 2022;Parsons et al., 2022). Indeed, deep learning has proven essential in making the wealth of available data beneficial for learning how to extract relevant features from the sensed information (Heaton et al., 2018). ...
Article
Full-text available
Underwater passive acoustics is used worldwide for multi-year monitoring of marine mammals. Yet, the large amount of audio recordings raises the need to automate the detection of acoustic events. For instance, the increasing number of Offshore Wind Farms (OWF) raises key environmental and societal issues relating to their impacts on wildlife. In this context, monitoring marine mammals along with information on their acoustic environment throughout the OWF life cycle is crucial. The objective of this study is to evaluate the ability of a single deep learning model to precisely detect and localize, in time and in frequency, the marine mammal sounds over a wide frequency range and classify them by species and sound types. A broadband hydrophone, deployed at the Fécamp OWF (Normandy, France), recorded the underwater soundscape including sounds from marine mammals occurring in the area. To visualize these sounds, 15-second spectrograms were computed. From these images, dolphin (D) and porpoise (P) sounds were manually annotated, including different types of sounds: Click-Trains (DCT, PCT), Buzzes (DB, PB) and Whistles (DW). The spectrograms were then split into five-fold cross-validation datasets, each containing one half of manual annotations and one half of only background noise. A Faster R-CNN model was trained to precisely detect and classify the marine mammal sounds in the spectrograms. Three model output configurations were used: (1) overall detection of marine mammals (presence vs. absence), (2) detection and classification of species (two classes: dolphin, porpoise) and (3) sound types (five classes: DCT, DB, DW, PCT, PB). For the simplest configuration (1) 15.4% of the spectrogram dataset had detections while missing only 6.6% of annotated spectrograms. For the more precise configurations, (2) and (3), the mean Average Precision (mAP) achieved were 92.3% (2) and 84.3% (3), and the macro average Area under the curve (AUC) 95.7% (2) and 94.9% (3). This model will help to speed up the annotation processes, by reducing the spectrogram quantity to be manually analyzed and having time-frequency boxes already drawn. Several model parameters can be adjusted to trade off missed detections and false positives which need to be carefully considered and adapted to the problem. For instance, these adjustments would be particularly relevant depending on the human resources available to manually check the model detections and the criticality of missing marine mammal sounds. These models are promising, ranging from the simple detection of marine mammal presence to precise ecological inferences over the long term.
... Unfortunately, the images obtained are generally noisy and require image processing techniques such as thresholding, edge detection and morphological operations [23,24]. However, these methods are often labor-intensive, time-consuming, and prone to inaccuracies due to the complex and dynamic nature of marine environments [25]. Maintenance further evolved toward intelligent maintenance, which aims to replace human interventions with systems based on artificial intelligence. ...
Article
Full-text available
Biofouling, the accumulation of marine organisms on submerged surfaces, presents significant operational challenges across various marine industries. Traditional detection methods are labor-intensive and costly, necessitating the development of automated systems for efficient monitoring. The study presented in this paper focuses on detecting biofouling on tidal stream turbine blades using camera-based surveillance. The process begins with dividing the video into a series of images, which are then annotated to identify and select the bounding boxes containing objects to be detected. These annotated images are used to train YOLO version 8 to detect biofouled and clean blades in the images. The proposed approach is evaluated using metrics that demonstrate the superiority of this YOLO version compared to previous ones. To address the issue of misdetection, a data augmentation approach is proposed and tested across different YOLO versions, showing its effectiveness in improving detection quality and robustness.
... Once the image enhancements were applied within this study, there was a 9.94% increase in useable images, and therefore data availability, for the training process of the final deep learning model. Data availability and representativeness are important parameters to consider when incorporating deep learning approaches into environmental studies as they will directly impact the accuracy of outputs [53]. This is crucial for environments which are challenging to survey, such as tidal stream environments, and therefore lack the representative datasets required to feed into automated data interpretation. ...
... CNN-based techniques also outperform other machine learning methods, by up to 30% in some cases, with the deep learning architecture being able to handle complex data patterns more efficiently and detect shapes that are difficult for humans to discern [53,63]. ...
... Deep learning has the capability to solve multifaceted tasks within marine research and will play a key role in increasing the efficiency of routine data processing and reducing the amount of manual work associated with large, complex, datasets [53]. However, there are still drawbacks to deep learning use and areas where further development is required. ...
Article
Full-text available
Tidal stream environments are important areas of marine habitat for the development of marine renewable energy (MRE) sources and as foraging hotspots for megafaunal species (seabirds and marine mammals). Hydrodynamic features can promote prey availability and foraging efficiency that influences megafaunal foraging success and behaviour, with the potential for animal interactions with MRE devices. Uncrewed aerial vehicles (UAVs) offer a novel tool for the fine-scale data collection of surface turbulence features and animals, which is not possible through other techniques, to provide information on the potential environmental impacts of anthropogenic developments. However, large imagery datasets are time-consuming to manually review and analyse. This study demonstrates an experimental methodology for the automated detection of turbulence features within UAV imagery. A deep learning architecture, specifically a Faster R-CNN model, was used to autonomously detect kolk-boils within UAV imagery of a tidal stream environment. The model was trained on pre-existing, labelled images of kolk-boils that were pre-treated using a suite of image enhancement techniques based on the environmental conditions present within each image. A 75-epoch model variant provided the highest average recall and precision values; however, it appeared to be limited by sub-optimal detections of false positive values. Although further development is required, including the creation of standardised image data pools, increased model benchmarking and the advancement of tailored pre-processing techniques, this work demonstrates the viability of utilising deep learning to automate the detection of surface turbulence features within a tidal stream environment.
... The current development of Artificial Intelligence (AI) applied to image processing is increasingly solving challenges such as underwater target detection (Malde et al., 2020). Applications in autonomous monitoring of marine ecosystems and their resources (e.g., stocks) include faunal classification and tracking by fixed and mobile robotic platforms (Aguzzi et al., 2020b;Beyan and Browman, 2020;Goodwin et al., 2022;Lopez-Marcano et al., 2021;Lopez-Vazquez et al., 2020). By processing time-stamped and geo-referenced images for detection, identification, and tracking of species on board each platform, automated routines can produce time series of animal counts in real time, hence allowing for more efficient data storage and transfer, while reducing facility spaces and bandwidth demands (Aguzzi et al., 2012;Aguzzi et al., 2020b;Aguzzi et al., 2022). ...
Article
Edge computing on mobile marine platform is paramount for automated ecological monitoring. The goal of demonstrating the computational feasibility of an Artificial Intelligence (AI)-powered camera for fully automated real-time species-classification on deep-sea crawler platforms was searched by running You-Only-Look-Once (YOLO) model on an edge computing device (NVIDIA Jetson Nano), to evaluate the achievable animal detection performances, execution time and power consumption, using all the available cores. We processed a total of 337 rotating video scans (~180 •), taken during approximately 4 months in 2022 at the methane hydrates site of Barkley Canyon (Vancouver Island; BC; Canada), focusing on three abundant species (i.e., Sablefish Anoplopoma fimbria, Hagfish Eptatretus stoutii, and Rockfish Sebastes spp.). The model was trained on 1926 manually annotated video frames and showed high detection test performances in terms of accuracy (0.98), precision (0.98), and recall (0.99). The trained model was then applied on 337 videos. In 288 videos we detected a total of 133 Sablefish, 31 Hagfish, and 321 Rockfish nearly in real-time (about 0.31 s/image) with very low power consumption (0.34 J/image). Our results have broad implications on intelligent ecological monitoring. Indeed, YOLO model can meet operational-autonomy criteria for fast image processing with limited computational and energy loads.