Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Underwater Object Detection Enhancement via Channel Stabilization

Muhammad Ali, Salman Khan

TL;DR

Underwater object detection is hindered by haze and blue-dominated color casts in marine imagery. The authors propose a lightweight Image Enhancement Module (IEM) paired with a Channel Stabilization mechanism, plus targeted augmentations, and evaluate this pipeline across Detectron2 backbones on the TrashCan-1.0 dataset, reporting notable gains for small objects and overall AP, including comparisons to Deformable DETR. Key findings show up to about 9.53 percentage points improvements in small-object AP on the instance version and strong performance of RetinaNet with IEM+CSM+Sharpening, with domain-generalization benefits observed on additional datasets. The approach offers a computationally efficient path to stronger underwater debris detection with potential for practical deployment in marine pollution monitoring, with code to be released.

Abstract

The complex marine environment exacerbates the challenges of object detection manifold. Marine trash endangers the aquatic ecosystem, presenting a persistent challenge. Accurate detection of marine deposits is crucial for mitigating this harm. Our work addresses underwater object detection by enhancing image quality and evaluating detection methods. We use Detectron2's backbone with various base models and configurations for this task. We propose a novel channel stabilization technique alongside a simplified image enhancement model to reduce haze and color cast in training images, improving multi-scale object detection. Following image processing, we test different Detectron2 backbones for optimal detection accuracy. Additionally, we apply a sharpening filter with augmentation techniques to highlight object profiles for easier recognition. Results are demonstrated on the TrashCan Dataset, both instance and material versions. The best-performing backbone method incorporates our channel stabilization and augmentation techniques. We also compare our Detectron2 detection results with the Deformable Transformer. In the instance version of TrashCan 1.0, our method achieves a 9.53% absolute increase in average precision for small objects and a 7% absolute gain in bounding box detection compared to the baseline. The code will be available on Code: https://github.com/aliman80/Underwater- Object-Detection-via-Channel-Stablization

Underwater Object Detection Enhancement via Channel Stabilization

TL;DR

Underwater object detection is hindered by haze and blue-dominated color casts in marine imagery. The authors propose a lightweight Image Enhancement Module (IEM) paired with a Channel Stabilization mechanism, plus targeted augmentations, and evaluate this pipeline across Detectron2 backbones on the TrashCan-1.0 dataset, reporting notable gains for small objects and overall AP, including comparisons to Deformable DETR. Key findings show up to about 9.53 percentage points improvements in small-object AP on the instance version and strong performance of RetinaNet with IEM+CSM+Sharpening, with domain-generalization benefits observed on additional datasets. The approach offers a computationally efficient path to stronger underwater debris detection with potential for practical deployment in marine pollution monitoring, with code to be released.

Abstract

The complex marine environment exacerbates the challenges of object detection manifold. Marine trash endangers the aquatic ecosystem, presenting a persistent challenge. Accurate detection of marine deposits is crucial for mitigating this harm. Our work addresses underwater object detection by enhancing image quality and evaluating detection methods. We use Detectron2's backbone with various base models and configurations for this task. We propose a novel channel stabilization technique alongside a simplified image enhancement model to reduce haze and color cast in training images, improving multi-scale object detection. Following image processing, we test different Detectron2 backbones for optimal detection accuracy. Additionally, we apply a sharpening filter with augmentation techniques to highlight object profiles for easier recognition. Results are demonstrated on the TrashCan Dataset, both instance and material versions. The best-performing backbone method incorporates our channel stabilization and augmentation techniques. We also compare our Detectron2 detection results with the Deformable Transformer. In the instance version of TrashCan 1.0, our method achieves a 9.53% absolute increase in average precision for small objects and a 7% absolute gain in bounding box detection compared to the baseline. The code will be available on Code: https://github.com/aliman80/Underwater- Object-Detection-via-Channel-Stablization
Paper Structure (17 sections, 6 equations, 11 figures, 3 tables)

This paper contains 17 sections, 6 equations, 11 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methods
  4. Overview
  5. PreProcessing
  6. Image Enhancement
  7. Channel Stabilization
  8. Augmentations
  9. Experiments
  10. Experimental Setup
  11. Implementation Details
  12. Ablation Study
  13. State-of-the-art Comparison
  14. Quantitative Results
  15. Qualitative Results
  16. ...and 2 more sections

Figures (11)

  • Figure 1: The Architecture of proposed method. In the first block pre-processing is done with image enhancement model then channel stabilization scheme is applied. Image is then processed in Detectron2 using various backbones: RetinaNet output is given in red block while for Faster R-CNN it follows flow given in grey and use ROI heads for final detction.
  • Figure 2: The Image Enhancement block and Channel Stabilization module which equalizes the image brightness and color restoration.
  • Figure 3: Channel Stabilized image.
  • Figure 4: Background removal.
  • Figure 5: Individual Class Detection for Retina Net - Material-version.
  • ...and 6 more figures