Mobile Detection Benchmark

This repo is used to test the speed of the mobile terminal models

Benchmark Result

Model	Input size	mAP^val 0.5:0.95	mAP^val 0.5	Params ^(M)	FLOPS ^(G)	Latency^{NCNN^(ms)}	Latency^{lite^(ms)}	Config
YOLOv3-Tiny	416*416	16.6	33.1	8.86	5.62	25.42	-	model link
YOLOv4-Tiny	416*416	21.7	40.2	6.06	6.96	23.69	-	model link
PP-YOLO-Tiny	320*320	20.6	-	1.08	0.58	6.75	-	model link
PP-YOLO-Tiny	416*416	22.7	-	1.08	1.02	10.48	-	model link
Nanodet-M	320*320	20.6	-	0.95	0.72	8.71	-	model link
Nanodet-M	416*416	23.5	-	0.95	1.2	13.35	-	model link
Nanodet-M 1.5x	416*416	26.8	-	2.08	2.42	15.83	-	model link
YOLOX-Nano	416*416	25.8	-	0.91	1.08	19.23	-	model link
YOLOX-Tiny	416*416	32.8	-	5.06	6.45	32.77	-	model link
YOLOv5n	640*640	28.4	46.0	1.9	4.5	40.35	-	model link
YOLOv5s	640*640	37.2	56.0	7.2	16.5	78.05	-	model link
PicoDet-S	320*320	27.1	41.4	0.99	0.73	8.13	6.65	model link
PicoDet-S	416*416	30.6	45.5	0.99	1.24	12.37	9.82	model link
PicoDet-M	320*320	30.9	45.7	2.15	1.48	11.27	9.61	model link
PicoDet-M	416*416	34.3	49.8	2.15	2.50	17.39	15.88	model link
PicoDet-L	320*320	32.6	47.9	3.24	2.18	15.26	13.42	model link
PicoDet-L	416*416	35.9	51.7	3.24	3.69	23.36	21.85	model link
PicoDet-L	640*640	40.3	57.1	3.24	8.74	54.11	50.55	model link
PicoDet-Shufflenetv2 1x	416*416	30.0	44.6	1.17	1.53	15.06	10.63	model link
PicoDet-MobileNetv3-large 1x	416*416	35.6	52.0	3.55	2.80	20.71	17.88	model link
PicoDet-LCNet 1.5x	416*416	36.3	52.2	3.10	3.85	21.29	20.8	model link

Table Notes:

Latency: All our models test on Qualcomm Snapdragon 865(4\*A77+4\*A55) with 4 threads by arm8 and with FP16. In the above table, test latency on NCNN and Lite->Paddle-Lite.
All model are trained on COCO train2017 dataset and evaluated on COCO val2017.