Efficient Marine Organism Detector (EMOD) for Marine Video Surveillance

We release the code of Efficient Marine Organism Detector (EMOD) in our papers:

Detecting Marine Organisms via Joint Attention-Relation Learning for Marine Video Surveillance. (IEEE Journal of Oceanic Engineering, 2022, DOI: 10.1109/JOE.2022.3162864 (DOI currently unavailable))
Detecting Organisms for Marine Video Surveillance. (Global OCEANS 2020, DOI: 10.1109/IEEECONF38699.2020.9389458)

@inproceedings{shi2020detecting,
  title={Detecting Organisms for Marine Video Surveillance},
  author={Shi, Zhensheng and Guan, Cheng and Cao, Liangjie and Li, Qianqian and Liang, Ju and Guo, Zonghui and Gu, Zhaorui and Zheng, Haiyong and Zheng, Bing},
  booktitle={Global OCEANS 2020: Singapore--US Gulf Coast},
  pages={1--7},
  year={2020},
  organization={IEEE}
}

Introduction

We design an Efficient Marine Organism Detector (EMOD) for high-resolution marine video surveillance to detect organisms and surveil marine environments in a real-time and fast fashion. We also propose a novel Attention-Relation (AR) module to explore joint Attention-Relation in CNNs for marine organism detection. This code is based on the mmdetection codebase (v2.13.0).

Requirements

Linux or macOS (Windows is in experimental support)
Python 3.6+
PyTorch 1.3+
CUDA 9.2+ (If you build PyTorch from source, CUDA 9.0 is also compatible)
GCC 5+
MMCV

Datasets

HabCam, MOUSS and MBARI.

Contact dataset provider, and download the datasets and annotations: CVPR 2018 Workshop or CVPR 2019 Workshop.
Put all images and annotation files to $EMOD/data folder.

Running

To train a FR-R50 detector on HABCAM dataset, you can run the script:
```
CUDA_VISIBLE_DEVICES=0,1 PORT=29500 tools/dist_train.sh configs-emod/FR/FR_R50_FPN_HABCAM_SP1.py 2
```
You can also set the variables (CONFIG_FILE, GPU_NUM) in scripts/run_habcam.sh, and then run the script:
```
bash scripts/run_habcam.sh
```

Models

We will provide the models and results later.

Acknowledgement

We really appreciate the contributors of following codebases.

open-mmlab/mmdetection

zhenglab / EMOD