YOLOMG

Codes and dataset for the paper "YOLOMG: Vision-based Drone-to-Drone Detection with Appearance and Pixel-Level Motion Fusion"

Dataset

ARD100 dataset

• BaiduYun (code:1x2z)

codes to generate dataset is in ./test_code directory

python generate_mask5.py is applied to generate mask32

python YOLOMG_extract_frames.py is used to generate images

python generate_dataset.py is used to generate train/test datasets. The videos id division is included.

data processing in ./data directory

dataset spliting

python3 split_train_val.py --xml_path xx/xxx/Annotations --txt_path xx/xxx/ImageSets/Main

transfer voc label to yolo label with .txt files

python3 voc2yolo.py

generate images directory, train.txt, val.txt, test.txt

python3 voc_label.py

generate mask directory, train2.txt, val2.txt, test2.txt

Python3 voc_label2.py

train

python3 train.py --data data/NPS.yaml --cfg models/NPS_uav_s.yaml --weights yolov5s.pt --batch-size 8 --epochs 100 --imgsz 1280 --name NPS-1280

val

python3 val.py --weights runs/train/NPS-1280/weights/best.pt --data data/NPS_test.yaml --task val --conf-thres 0.001 --name NPS_test-1280 --imgsz 1280 --batch-size 8 --device 0

fast demo test

python3 dualdetector.py

DDP train

python -m torch.distributed.run --nproc_per_node=4 --master_port 12345 train.py --data data/ARD100_mask32.yaml --cfg models/ARD100_drone_s.yaml --weights yolov5s.pt --batch-size 16 --epochs 100 --imgsz 1280 --name ARD100_mask32-1280 --device 0,1,2,3