Skip to content

Latest commit

 

History

History
60 lines (41 loc) · 1.61 KB

README.md

File metadata and controls

60 lines (41 loc) · 1.61 KB
Raw

YOLOv5-ODConvNeXt

This repo is the implementation for Deep learning based efficient ship detection from drone-captured images for maritime surveillance - ScienceDirect.

YOLOv5-ODConvNeXt is an improved version of YOLOv5s for ship detection on drone-captured images.

Install

git clone https://github.com/chengshuxiao/YOLOv5-ODConvNeXt.git  # clone
cd YOLOv5-ODConvNeXt
pip install -r requirements.txt  # install

Inference with detect.py

A trained YOLOv5-ODConvNeXt model is provided in checkpoints/yolov5-odconvnext.pt，the detection results are saved in runs/detect

python detect.py
    --weights checkpoints/yolov5-odconvnext.pt
    --source data/images
    --line-thickness 2

Two demo outputs are shown below.

Train

The configuration of our model is in models/yolov5-odconvnext.yaml

python train.py
    --data data/MyShip3200.yaml
    --cfg models/yolov5-odconvnext.yaml
    --hyp data/hyps/hyp.scratch-low.yaml
    --epoch 500
    --batch-size 32
    --device 0
    --workers 8

Dataset

Link: https://pan.baidu.com/s/1CeNyZZVp6RLDbIEdBikmGQ

Access Key: cuc6

References

Thanks to their great works.