SeeCucumbers: Using Deep Learning and Drone Imagery to Detect Sea Cucumbers on Coral Reef Flats

Abstract

Sea cucumbers (Holothuroidea or holothurians) are a valuable fishery and are also crucial nutrient recyclers, bioturbation agents, and hosts for many biotic associates. Their ecological impacts could be substantial given their high abundance in some reef locations and thus monitoring their populations and spatial distribution is of research interest. Traditional in situ surveys are laborious and only cover small areas but drones offer an opportunity to scale observations more broadly, especially if the holothurians can be automatically detected in drone imagery using deep learning algorithms. We adapted the object detection algorithm YOLOv3 to detect holothurians from drone imagery at Hideaway Bay, Queensland, Australia. We successfully detected 11,462 of 12,956 individuals over 2.7ha with an average density of 0.5 individual/m2. We tested a range of hyperparameters to determine the optimal detector performance and achieved 0.855 mAP, 0.82 precision, 0.83 recall, and 0.82 F1 score. We found as few as ten labelled drone images was sufficient to train an acceptable detection model (0.799 mAP). Our results illustrate the potential of using small, affordable drones with direct implementation of open-source object detection models to survey holothurians and other shallow water sessile species.

Survey of freshwater mussels using high‐resolution acoustic imaging sonar and deep learning‐based object detection in Lake Izunuma, Japan

“…The proposed acoustic survey system is limited in its ability to distinguish between different mussel species. This is similar to some of the issues encountered when remote sensing techniques are used to survey benthic organisms (Li et al, 2021). Therefore, supplementary quadrat surveys by divers may help provide a more comprehensive understanding of the ecological and biological roles of specific species in locations with multiple species.…”

“…A confusion matrix is formed by integrating ground truth data with the outcomes identified by a deep learning model. The metrics for evaluating mussel detection were computed and interpreted following the guidelines given in Table 2 (Everingham et al, 2010; Everingham & Winn, 2011; Lin et al, 2014; Géron, 2019; Li et al, 2021).…”

Antarctic Seabed Assemblages in an Ice-Shelf-Adjacent Polynya, Western Weddell Sea

“…In the present study, deposit-feeding crawlers, all of which were holothurians, were found to be the strongest surrogacy candidate for overall megafaunal density ( Figure 6 ). Although preliminary, this finding could be especially advantageous for imagery-based seafloor assessments given that holothurians are relatively easy to detect and identify due to their visual characteristics and mobility, making them also potentially ideal candidates for the development of automated appraisal and monitoring tools (e.g., [ 145 ]). Nevertheless, it remains important to continue testing proposed surrogates.…”

The unique value proposition for using drones to map coastal ecosystems

“…However, drones have allowed researchers to more accurately map these ecosystems with increased spatial resolution and explore how biotic and abiotic environmental variability influences classification (Nahirnick et al, 2019). Augmented by the increased spatial resolution drones provide, advanced classification techniques, such as machine and deep learning also add enhanced contextual information (Hamylton et al, 2020a), including mapping down to the individual organism level in sea cucumber populations (Li et al, 2021). When increasing spatial resolution alone is not enough to map and monitor heterogenous coastal ecosystem features, advanced data types such as hyperspectral (Cornet and Joyce, 2021;Jaud et al, 2021) and Light Detection and Ranging (LiDAR) (Kramer et al, 2021) can be employed and are especially useful on drones due to their cost reduction and advances in sensor miniaturisation in current technology.…”