APSIPA Transactions on Signal and Information Processing > Vol 13 > Issue 1

GreenCOD: A Green Camouflaged Object Detection Method

Hong-Shuo Chen, University of Southern California, USA, hongshuo@usc.edu , Yao Zhu, University of Southern California, USA, Suya You, DEVCOM Army Research Laboratory, USA, Azad M. Madni, University of Southern California, USA, C.-C. Jay Kuo, University of Southern California, USA
 
Suggested Citation
Hong-Shuo Chen, Yao Zhu, Suya You, Azad M. Madni and C.-C. Jay Kuo (2024), "GreenCOD: A Green Camouflaged Object Detection Method", APSIPA Transactions on Signal and Information Processing: Vol. 13: No. 1, e31. http://dx.doi.org/10.1561/116.20240009

Publication Date: 02 Dec 2024
© 2024 H.-S. Chen, Y. Zhu, S. You, A. M. Madni and C.-C. J. Kuo
 
Subjects
Image and video processing,  Statistical signal processing,  Statistical/Machine learning,  Segmentation and grouping,  Learning and statistical methods
 
Keywords
Green LearningExtreme Gradient Boosting (XGBoost)Camouflaged Object Detection
 

Share

Open Access

This is published under the terms of CC BY-NC.

Downloaded: 13 times

In this article:
Introduction 
Related Work 
GreenCOD Method 
Experiments 
Conclusion and Future Work 
Acknowledgments 
References 

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. Traditional camouflaged object detection approaches rely on complex deep neural networks, seeking performance improvements by backpropagation-based finetuning. However, such methods are typically computationally demanding and exhibit only marginal performance variations across different models. It raises the question of whether effective training can be achieved without backpropagation. In this direction, 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. This new, more efficient paradigm opens avenues for further exploration in green, backpropagation-free model training. We make GreenCOD source code and on-device demo available at https://greencod.ai/ for futher research.

DOI:10.1561/116.20240009