YOLOv7t-CEBC Network for Underwater Litter Detection

Xinyu Zhang,Daqi Zhu,Wenyang Gan

DOI: https://doi.org/10.3390/jmse12040524

IF: 2.744

2024-03-23

Journal of Marine Science and Engineering

Abstract:The issue of marine litter has been an important concern for marine environmental protection for a long time, especially underwater litter. It is not only challenging to clean up, but its prolonged presence underwater can cause damage to marine ecosystems and biodiversity. This has led to underwater robots equipped with powerful visual detection algorithms becoming the mainstream alternative to human labor for cleaning up underwater litter. This study proposes an enhanced underwater litter detection algorithm, YOLOv7t-CEBC, based on YOLOv7-tiny, to assist underwater robots in target identification. The research introduces some modules tailored for marine litter detection within the model framework, addressing inter-class similarity and intra-class variability inherent in underwater waste while balancing detection precision and speed. Experimental results demonstrate that, on the Deep Plastic public dataset, YOLOv7t-CEBC achieves a detection accuracy (mAP) of 81.8%, markedly surpassing common object detection algorithms. Moreover, the detection frame rate reaches 118 FPS, meeting the operational requirements of underwater robots. The findings affirm that the enhanced YOLOv7t-CEBC network serves as a reliable tool for underwater debris detection, contributing to the maintenance of marine health.

oceanography,engineering, marine, ocean

What problem does this paper attempt to address?

The paper primarily addresses the issue of underwater garbage detection in marine environments and proposes an improved object detection algorithm—the YOLOv7t-CEBC network. Specifically, the paper tackles the following key issues:

1. **Marine Garbage Pollution Problem**: The large amount of plastic waste in the ocean has severely impacted the ecological environment, making the cleanup of this waste particularly important.

2. **Limitations of Existing Solutions**: Current solutions include manual cleanup (which is inefficient, costly, and poses safety risks) and the use of intelligent devices such as underwater robots for cleanup. Although the latter is more efficient and safer, it still faces challenges in practical applications, especially in the performance of visual detection algorithms in complex underwater environments.

3. **Improving Underwater Garbage Detection Performance**: To enhance the detection capabilities of underwater robots for garbage, this paper proposes an improved version of YOLOv7-tiny, the YOLOv7t-CEBC network. This network enhances detection performance by introducing several modules specifically designed for marine garbage, particularly addressing the issues of similarity between different categories of garbage and variability within the same category.

4. **Algorithm Innovations**:
- Introduced ConvNeXt Block (CNeB) in the network backbone, utilizing its efficient and simple fully convolutional structure to enhance feature extraction capabilities.
- Introduced EMA (Efficient Multi-scale Attention Mechanism) and Biformer (Biformer Attention Mechanism), which respectively improve the ability to capture global object features and the detection performance in densely populated small object areas.
- Replaced the upsampling layer with CARAFE (Content-Aware ReAssembly of FEatures) operation, providing a larger receptive field for feature reassembly and guiding the reassembly process based on input information to improve performance.

5. **Experimental Validation**: Experiments were conducted on the publicly available Deep Plastic dataset, and the results showed that the YOLOv7t-CEBC network achieved a detection accuracy (mAP) of 81.8% and a frame rate of 118 FPS, demonstrating the effectiveness and practicality of the network in underwater garbage detection.

In summary, this research aims to address key issues in marine environmental protection by proposing a high-performance underwater garbage detection algorithm and providing technical support for the garbage cleanup work of underwater robots.