Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Oil Spill Segmentation using Deep Encoder-Decoder models

Abhishek Ramanathapura Satyanarayana, Maruf A. Dhali

TL;DR

This work investigates end-to-end oil-spill detection in SAR imagery using deep encoder–decoder segmentation models. By evaluating multiple encoders (ResNet and EfficientNet) with DeepLabV3/DeepLabV3+ decoders on the OilSpillDataset, it demonstrates that a ResNet-50 encoder with DeepLabV3+ achieves the best test-set mean IoU of $64.868\%$ and an oil-spill class IoU of $61.549\%$, while outperforming previous patch-based baselines in class accuracy. The study shows that high-dimensional, unpatched SAR inputs can approach or exceed previous patch-based performance and highlights class-confusion challenges between oil spill and look-alike classes. It also points to potential enhancements such as visual self-attention to further improve detection fidelity in operational settings.

Abstract

Crude oil is an integral component of the world economy and transportation sectors. With the growing demand for crude oil due to its widespread applications, accidental oil spills are unfortunate yet unavoidable. Even though oil spills are difficult to clean up, the first and foremost challenge is to detect them. In this research, the authors test the feasibility of deep encoder-decoder models that can be trained effectively to detect oil spills remotely. The work examines and compares the results from several segmentation models on high dimensional satellite Synthetic Aperture Radar (SAR) image data to pave the way for further in-depth research. Multiple combinations of models are used to run the experiments. The best-performing model is the one with the ResNet-50 encoder and DeepLabV3+ decoder. It achieves a mean Intersection over Union (IoU) of 64.868% and an improved class IoU of 61.549% for the ``oil spill" class when compared with the previous benchmark model, which achieved a mean IoU of 65.05% and a class IoU of 53.38% for the ``oil spill" class.

Oil Spill Segmentation using Deep Encoder-Decoder models

TL;DR

This work investigates end-to-end oil-spill detection in SAR imagery using deep encoder–decoder segmentation models. By evaluating multiple encoders (ResNet and EfficientNet) with DeepLabV3/DeepLabV3+ decoders on the OilSpillDataset, it demonstrates that a ResNet-50 encoder with DeepLabV3+ achieves the best test-set mean IoU of and an oil-spill class IoU of , while outperforming previous patch-based baselines in class accuracy. The study shows that high-dimensional, unpatched SAR inputs can approach or exceed previous patch-based performance and highlights class-confusion challenges between oil spill and look-alike classes. It also points to potential enhancements such as visual self-attention to further improve detection fidelity in operational settings.

Abstract

Crude oil is an integral component of the world economy and transportation sectors. With the growing demand for crude oil due to its widespread applications, accidental oil spills are unfortunate yet unavoidable. Even though oil spills are difficult to clean up, the first and foremost challenge is to detect them. In this research, the authors test the feasibility of deep encoder-decoder models that can be trained effectively to detect oil spills remotely. The work examines and compares the results from several segmentation models on high dimensional satellite Synthetic Aperture Radar (SAR) image data to pave the way for further in-depth research. Multiple combinations of models are used to run the experiments. The best-performing model is the one with the ResNet-50 encoder and DeepLabV3+ decoder. It achieves a mean Intersection over Union (IoU) of 64.868% and an improved class IoU of 61.549% for the ``oil spill" class when compared with the previous benchmark model, which achieved a mean IoU of 65.05% and a class IoU of 53.38% for the ``oil spill" class.
Paper Structure (17 sections, 5 equations, 6 figures, 4 tables)

This paper contains 17 sections, 5 equations, 6 figures, 4 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORKS
  3. METHODOLOGY
  4. Dataset
  5. Image Preprocessing & Data Augmentation
  6. Models
  7. Residual networks
  8. EfficientNet
  9. DeepLabV3 & DeepLabV3+
  10. Training Models
  11. Transfer Learning
  12. Evaluation Metrics
  13. RESULTS & DISCUSSION
  14. Quantitative Analysis
  15. Qualitative Analysis
  16. ...and 2 more sections

Figures (6)

  • Figure 1: Plot showing the distribution of the semantic classes in the Oil Spill Detection Dataset.
  • Figure 2: (a) Example of an original image from the training set. (b) A sample padded image from the training set.
  • Figure 3: (a) Sample test image (b) Groundtruth (c) Predicted mask.
  • Figure 4: (a) Sample test image (b) Groundtruth (c) Predicted mask.
  • Figure 5: (a) Sample test image (b) Groundtruth (c) Predicted mask.
  • ...and 1 more figures