model 성능표

[Batwan01]

📄 Describe

Model의 성능을 기록하여 앙상블, 활용에 도움을 주고자 함

• 모델 성능표

model	batch	lr	epoch	Val mAP 50	Public mAP50
yolov5	8	0.005	1	1
DINO	1	0.005	1	0.519
CO-DETR(Swin-L, lsj)	1	0.00002	3	0.513
CO-DETR(Swin-L, Object365)	1	0.00001	1	0.64	0.6533
CO-DETR(Swin-L, Object365)	1	2e-07	3	0.683	0.688
CO-DETR(Obj365, 2048, oversampling)	1	0.00001	1	0.7010
Cascade-RCNN(Swin-L, 2048)	1	1e-05	1	0.647
Cascade-RCNN(Swin-L, 2048)	1	1e-05	2	0.653
Cascade-RCNN(Swin-L, 2048)	1	1e-06	3	0.6720
Cascade-RCNN(Swin-L, 2048)	1	1e-06	4	0.6690
Cascade-RCNN(Swin-L, 2048)	1	1e-06	5	0.6830
Cascade-RCNN(Swin-L, 2048, oversampling)	1	1e-06	1	0.6790
Cascade-RCNN(Swin-L, 2048, oversampling)	1	1e-06	2	0.6880
Cascade-RCNN(Swin-L, 4096)			2	0.604

• 모델 학습 로그

  • CO-DETR(Swin-L, lsj)

    Epoch	coco/bbox_mAP	coco/bbox_mAP_50	coco/bbox_mAP_75	coco/bbox_mAP_s	coco/bbox_mAP_m	coco/bbox_mAP_l	data_time	time
    1	0.271	0.427	0.301	0.213	0.522	0.511	0.0019	0.3132
    2	0.311	0.509	0.351	0.275	0.477	0.539	0.0020	0.3123
    3	0.324	0.513	0.363	0.284	0.537	0.599	0.0020	0.3122

🙋🏻 More