Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GreenCOD: A Green Camouflaged Object Detection Method

Hong-Shuo Chen, Yao Zhu, Suya You, Azad M. Madni, C. -C. Jay Kuo

TL;DR

GreenCOD tackles camouflaged object detection with a green, backpropagation-free approach that fuses pre-trained deep features with gradient-boosted trees. By replacing end-to-end training with a four-stage, multi-scale XGBoost pipeline and Neighborhood Construction, it achieves competitive COD performance while significantly reducing MACs and parameters. The method demonstrates state-of-the-art efficiency on COD10K under 50G MACs and substantial MACs reductions versus larger models on NC4K, while maintaining interpretability and avoiding full network fine-tuning. Overall, the paper introduces a viable green learning paradigm for COD and outlines promising future directions such as lighter feature extractors and broader applications to video COD and edge detection.

Abstract

We introduce GreenCOD, a green method for detecting camouflaged objects, distinct in its avoidance of backpropagation techniques. GreenCOD leverages gradient boosting and deep features extracted from pre-trained Deep Neural Networks (DNNs). Traditional camouflaged object detection (COD) approaches often rely on complex deep neural network architectures, seeking performance improvements through backpropagation-based fine-tuning. However, such methods are typically computationally demanding and exhibit only marginal performance variations across different models. This raises the question of whether effective training can be achieved without backpropagation. Addressing this, our work proposes a new paradigm that utilizes gradient boosting for COD. This approach significantly simplifies the model design, resulting in a system that requires fewer parameters and operations and maintains high performance compared to state-of-the-art deep learning models. Remarkably, our models are trained without backpropagation and achieve the best performance with fewer than 20G Multiply-Accumulate Operations (MACs). This new, more efficient paradigm opens avenues for further exploration in green, backpropagation-free model training.

GreenCOD: A Green Camouflaged Object Detection Method

TL;DR

GreenCOD tackles camouflaged object detection with a green, backpropagation-free approach that fuses pre-trained deep features with gradient-boosted trees. By replacing end-to-end training with a four-stage, multi-scale XGBoost pipeline and Neighborhood Construction, it achieves competitive COD performance while significantly reducing MACs and parameters. The method demonstrates state-of-the-art efficiency on COD10K under 50G MACs and substantial MACs reductions versus larger models on NC4K, while maintaining interpretability and avoiding full network fine-tuning. Overall, the paper introduces a viable green learning paradigm for COD and outlines promising future directions such as lighter feature extractors and broader applications to video COD and edge detection.

Abstract

We introduce GreenCOD, a green method for detecting camouflaged objects, distinct in its avoidance of backpropagation techniques. GreenCOD leverages gradient boosting and deep features extracted from pre-trained Deep Neural Networks (DNNs). Traditional camouflaged object detection (COD) approaches often rely on complex deep neural network architectures, seeking performance improvements through backpropagation-based fine-tuning. However, such methods are typically computationally demanding and exhibit only marginal performance variations across different models. This raises the question of whether effective training can be achieved without backpropagation. Addressing this, our work proposes a new paradigm that utilizes gradient boosting for COD. This approach significantly simplifies the model design, resulting in a system that requires fewer parameters and operations and maintains high performance compared to state-of-the-art deep learning models. Remarkably, our models are trained without backpropagation and achieve the best performance with fewer than 20G Multiply-Accumulate Operations (MACs). This new, more efficient paradigm opens avenues for further exploration in green, backpropagation-free model training.
Paper Structure (20 sections, 3 equations, 3 figures, 3 tables)

This paper contains 20 sections, 3 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Recent Approaches in COD
  4. Green Learning
  5. GreenCOD Method
  6. Feature Extraction
  7. Concatenation and Resizing
  8. Multi-scale XGBoost
  9. Neighborhood Construction (NC)
  10. Experiments
  11. Datasets
  12. Evaluation Metrics
  13. Experiment results
  14. Visualization analysis
  15. Ablation Study
  16. ...and 5 more sections

Figures (3)

  • Figure 1: An overview of the GreenCOD method, where the input is an image of dimension $672 \times 672 \times 3$, and the output is a probability mask of dimension $168 \times 168 \times 1$. NC stands for Neighborhood Construction.
  • Figure 2: Illustration of mask predictions using the proposed GreenCOD. Images are taken from the COD10K test dataset. From left to right: (a) tampered images, (b) ground-truth masks, (c) prediction.
  • Figure 3: The illustration of the prediction of each XGBoost