U-DECN: End-to-End Underwater Object Detection ConvNet with Improved DeNoising Training

Introduction

The official implementation of U-DECN: End-to-End Underwater Object Detection ConvNet with Improved DeNoising Training.

Installation

This project is based on MMDetection.

• Python 3.8
• PyTorch 1.13.1+cu116

Step 1. Create a conda virtual environment and activate it.

conda create -n udecn python=3.8 -y
conda activate udecn

Step 2. Install PyTorch following official instructions.

Linux and Windows

# Wheel CUDA 11.6
pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116

# Conda CUDA 11.6
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.6 -c pytorch -c nvidia

Step 3. Install MMDetection and dependent packages.

pip install -U openmim
mim install mmengine==0.10.1
mim install mmcv==2.0.1
mim install mmdet==3.2.0
pip install -r requirements.txt

Step 4. Build Deformable Convolutional Networks (DCN) CUDA operations.

# Build DCN CUDA ops
sh tools_build_dcn/make_dcn.sh

# Test DCN CUDA ops
sh tools_build_dcn/test_dcn.sh

Dataset

The data structure DUO looks like below:

# DUO

data
├── DUO
│   ├── annotations
│   │   ├── instances_train.json
│   │   ├── instances_test.json
│   ├── images
│   │   ├── train
│   │   ├── test

Usage

Training

# U-DECN 50 epoch
bash tools/dist_train.sh configs/u_decn/u-decn_r50_2xb4-50e_duo.py 2

# U-DECN 72 epoch
bash tools/dist_train.sh configs/u_decn/u-decn_r50_2xb4-6x_duo.py 2

# U-DECN 100 epoch
bash tools/dist_train.sh configs/u_decn/u-decn_r50_2xb4-100e_duo.py 2

Test

The weight .pth of U-DECN is available here.

# U-DECN 50 epoch
bash tools/dist_test.sh configs/u_decn/u-decn_r50_2xb4-50e_duo.py u-decn_r50_2xb4-50e_duo.pth 2

# U-DECN 72 epoch
bash tools/dist_test.sh configs/u_decn/u-decn_r50_2xb4-6x_duo.py u-decn_r50_2xb4-6x_duo.pth 2

# U-DECN 100 epoch
bash tools/dist_test.sh configs/u_decn/u-decn_r50_2xb4-100e_duo.py u-decn_r50_2xb4-100e_duo.pth 2

Cite

@article{liu2025udecn,
  title={U-DECN: End-to-End Underwater Object Detection ConvNet With Improved Denoising Training}, 
  author={Liu, Zhuoyan and Wang, Bo and Wang, Bing and Li, Ye},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  volume={63},
  pages={1-9},
  year={2025},
  doi={10.1109/TGRS.2025.3595158},
}