FCOS: Fully Convolutional One-Stage Object Detector

Train FCOS

Taking training FCOS_R18_1x on COCO as the example,

python main.py --cuda -d coco --root path/to/coco -m fcos_r18_1x --batch_size 16 --eval_epoch 2

Taking training FCOS_R18_1x on COCO as the example,

python -m torch.distributed.run --nproc_per_node=8 train.py --cuda -dist -d coco --root path/to/coco -m fcos_r18_1x --batch_size 16 --eval_epoch 2

Taking testing FCOS_R18_1x on COCO-val as the example,

python test.py --cuda -d coco --root path/to/coco -m fcos_r18_1x --weight path/to/fcos_r18_1x.pth -vt 0.4 --show

Taking evaluating FCOS_R18_1x on COCO-val as the example,

python main.py --cuda -d coco --root path/to/coco -m fcos_r18_1x --resume path/to/fcos_r18_1x.pth --eval_first

python demo.py --mode image --path_to_img path/to/image_dirs/ --cuda -m fcos_r18_1x --weight path/to/weight -vt 0.4 --show

python demo.py --mode video --path_to_vid path/to/video --cuda -m fcos_r18_1x --weight path/to/weight -vt 0.4 --show --gif

python demo.py --mode camera --cuda -m fcos_r18_1x --weight path/to/weight -vt 0.4 --show --gif

Model	scale	FPS^{FP32 RTX 4060}	AP^val 0.5:0.95	AP^val 0.5	Weight	Logs
FCOS_R18_1x	800,1333	24	34.0	52.2	ckpt	log
FCOS_R50_1x	800,1333	9	39.0	58.3	ckpt	log
FCOS_RT_R18_3x	512,736	56	35.8	53.3	ckpt	log
FCOS_RT_R50_3x	512,736	34	40.7	59.3	ckpt	log