Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

UDEEP: Edge-based Computer Vision for In-Situ Underwater Crayfish and Plastic Detection

Dennis Monari, Farhad Fassihi Tash, Jordan J. Bird, Ahmad Lotfi, Isibor Kennedy Ihianle, Salisu Wada Yahaya, Isibor Kennedy Ihianle, Md Mahmudul Hasan, Pedro Sousa, Pedro Machado

TL;DR

This work tackles the problem of monitoring invasive signal crayfish and underwater plastic debris by deploying edge-based computer vision on the Cognitive Edge Device (CED). It combines two public underwater datasets with four YOLO variants (yolov5n, yolov5s, yolov8n, yolov8s) evaluated on an NVIDIA Jetson NJON, revealing a trade-off between accuracy and edge efficiency, with YOLOv5s delivering the best mAP@0.5 ($0.90$) and precision ($0.93$) and YOLOv8n offering the fastest, most power-efficient edge inference ($0.29$ s; $0.09$ J on GPU). The methodology includes dataset pre-processing, extensive augmentation, and an edge-centric pipeline that integrates sensing data for robust, real-time monitoring. The results demonstrate that GPU-accelerated edge inference substantially outperforms CPU in both speed and energy, enabling practical deployment in remote aquatic environments. Looking forward, the paper proposes pose estimation and expansion to additional invasive species to enhance real-time conservation monitoring and policy support.

Abstract

Invasive signal crayfish have a detrimental impact on ecosystems. They spread the fungal-type crayfish plague disease (Aphanomyces astaci) that is lethal to the native white clawed crayfish, the only native crayfish species in Britain. Invasive signal crayfish extensively burrow, causing habitat destruction, erosion of river banks and adverse changes in water quality, while also competing with native species for resources leading to declines in native populations. Moreover, pollution exacerbates the vulnerability of White-clawed crayfish, with their populations declining by over 90%. To safeguard aquatic ecosystems, it is imperative to address the challenges posed by invasive species and pollution in aquatic ecosystem's. This article introduces the Cognitive Edge Device (CED) computing platform for the detection of crayfish and plastic. It also presents two publicly available underwater datasets, annotated with sequences of crayfish and aquatic plastic debris. Four You Only Look Once (YOLO) variants were trained and evaluated for crayfish and plastic object detection. YOLOv5s achieved the highest detection accuracy, with an mAP@0.5 of 0.90, and achieved the best precision

UDEEP: Edge-based Computer Vision for In-Situ Underwater Crayfish and Plastic Detection

TL;DR

This work tackles the problem of monitoring invasive signal crayfish and underwater plastic debris by deploying edge-based computer vision on the Cognitive Edge Device (CED). It combines two public underwater datasets with four YOLO variants (yolov5n, yolov5s, yolov8n, yolov8s) evaluated on an NVIDIA Jetson NJON, revealing a trade-off between accuracy and edge efficiency, with YOLOv5s delivering the best mAP@0.5 () and precision () and YOLOv8n offering the fastest, most power-efficient edge inference ( s; J on GPU). The methodology includes dataset pre-processing, extensive augmentation, and an edge-centric pipeline that integrates sensing data for robust, real-time monitoring. The results demonstrate that GPU-accelerated edge inference substantially outperforms CPU in both speed and energy, enabling practical deployment in remote aquatic environments. Looking forward, the paper proposes pose estimation and expansion to additional invasive species to enhance real-time conservation monitoring and policy support.

Abstract

Invasive signal crayfish have a detrimental impact on ecosystems. They spread the fungal-type crayfish plague disease (Aphanomyces astaci) that is lethal to the native white clawed crayfish, the only native crayfish species in Britain. Invasive signal crayfish extensively burrow, causing habitat destruction, erosion of river banks and adverse changes in water quality, while also competing with native species for resources leading to declines in native populations. Moreover, pollution exacerbates the vulnerability of White-clawed crayfish, with their populations declining by over 90%. To safeguard aquatic ecosystems, it is imperative to address the challenges posed by invasive species and pollution in aquatic ecosystem's. This article introduces the Cognitive Edge Device (CED) computing platform for the detection of crayfish and plastic. It also presents two publicly available underwater datasets, annotated with sequences of crayfish and aquatic plastic debris. Four You Only Look Once (YOLO) variants were trained and evaluated for crayfish and plastic object detection. YOLOv5s achieved the highest detection accuracy, with an mAP@0.5 of 0.90, and achieved the best precision
Paper Structure (19 sections, 6 equations, 7 figures, 3 tables)

This paper contains 19 sections, 6 equations, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. The challenges of underwater monitoring
  4. Early advances in underwater bio-diversity monitoring
  5. Modern application of computer vision for underwater bio-diversity monitoring
  6. Methodology
  7. ced prototype overview
  8. Network Architecture
  9. Datasets Pre-processing
  10. Dataset Augmentation
  11. Result Analysis
  12. Training Results
  13. Convergence analysis
  14. Model performance and power consumption on the njon
  15. Inference Time
  16. ...and 4 more sections

Figures (7)

  • Figure 1: ced monitoring workflow - end-to-end process of underwater monitoring. The system begins by recording underwater videos, followed by data import where camera and sensor metadata are stored and video frames extracted. The detection and tracking module then performs real-time identification of Signal Crayfish and plastic waste across frames. Finally, the data archiver formats metadata such as timestamps, geo-location, and depth, storing both detection and sensor data for later export and analysis, enabling biodiversity monitoring and pollution assessment.
  • Figure 2: A compilation showcasing a variety of underwater images that are included in the underwater plastic and signal crayfish dataset. A) A clear view of a sandy or gravelly bottom. There's a piece of decaying plastic material. B) A large, brownish plastic material in deeper or murkier water, obscuring the view causing limited visibility of the surrounding environment. C) A partially submerged plastic debris D) A close-up shot of a signal crayfish being held by a hand with claws and legs clearly visible. E) A view of a sandy or gravelly bottom with a juvenile signal crayfish partially visible. F) Numerous juvenile signal crayfish swimming amongst rocks and shells on the bottom.
  • Figure 3: pr curves for yolov5n, yolov5s, yolov8n and yolov8s models on crayfish and plastic detection. The light blue and orange curves represent the crayfish and plastic classes while the bold blue curve indicates the overall map at IoU 0.5. yolov5s achieves the highest map (0.900), followed closely by yolov8s (0.887).
  • Figure 4: F1-score vs. confidence threshold curves for yolov5n, yolov5s, yolov8n, yolov8s models. The light blue and orange curves represent the crayfish and plastic classes while the bold blue curve indicates the overall F1-score. yolov8s achieves the highest F1-score of 0.87 at a confidence threshold of 0.626, followed closely by yolov8n 0.85 at a confidence threshold of 0.541.
  • Figure 5: Convergence analysis of yolov5s, yolov5n, yolov8s, and yolov8n over 580 training epochs. The top plot shows mAP@0.5 progression, where yolov5s converges fastest and achieves the highest detection accuracy. The bottom plot shows validation box loss, with yolov8s exhibiting the lowest and most stable loss, indicating superior localization performance. Overall, yolov8s demonstrates the best convergence speed and training stability among the evaluated models.
  • ...and 2 more figures