UniRepLKNet Family

Paper | GitHub

axis

Select an option

Results

Parameters (M)

Images / Second

Publication Date

task

Select an option

---------

Object Detection

Instance Segmentation

Classification

Semantic Segmentation

Panoptic Segmentation

dataset

Select an option

---------

COCO (val)

COCO (test)

Cityscapes (val)

Cityscapes (test)

ADE20K (val)

ADE20K (test)

metric

Select an option

----------

mAP

AP₅₀

AP₇₅

mAP_s

mAP_m

mAP_l

GFLOPs

head

Select an option

---------

Mask R-CNN

Cascade Mask R-CNN

Mask2Former

HTC++

HTC

Panoptic FPN

axis

Select an option

Results

Parameters (M)

Images / Second

GFLOPs

Publication Date

Filters

pretrain dataset

Select an option

---------

JFT-3B

ImageNet-1k

ImageNet-22k

JFT-300M

pretrain method

Select an option

----------

Supervised

FCMAE

MAE

Legend

Legend attribute

Select an option

Family

Pretrain Dataset

Pretrain Method

Instance Head

Instance Training Epochs

Legend attribute (second)

Select an option

----------

Family

Pretrain Method

Instance Head

Instance Training Epochs

model	params (m)	pretrain	head	train	GFLOPs	mAP
UniRepLKNet-T	31.0	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	749.0	44.9
UniRepLKNet-S	55.6	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	835.0	45.9
UniRepLKNet-S	55.6	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	835.0	47.1
UniRepLKNet-B	97.9	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	978.0	47.4
UniRepLKNet-L	218.3	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1385.0	48.4
UniRepLKNet-XL	386.4	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1952.0	49.0

model	params (m)	pretrain	finetune	gflops	IN-1k
UniRepLKNet-A	4.4	IN-1k : Sup. : 300	— : — : —	0.6	77.0/—
UniRepLKNet-F	6.2	IN-1k : Sup. : 300	— : — : —	0.9	78.6/—
UniRepLKNet-P	10.7	IN-1k : Sup. : 300	— : — : —	1.6	80.2/—
UniRepLKNet-N	18.3	IN-1k : Sup. : 300	— : — : —	2.8	81.6/—
UniRepLKNet-T	31.0	IN-1k : Sup. : 300	— : — : —	4.9	83.2/—
UniRepLKNet-S	55.6	IN-1k : Sup. : 300	— : — : —	9.1	83.9/—
UniRepLKNet-S	55.6	IN-22k : Sup. : 90	IN-1k : 30 : 384	26.7	86.4/—
UniRepLKNet-B	97.9	IN-22k : Sup. : 90	IN-1k : 30 : 384	47.2	87.4/—
UniRepLKNet-L	218.3	IN-22k : Sup. : 90	IN-1k : 20 : 384	105.4	87.9/—
UniRepLKNet-XL	386.4	IN-22k : Sup. : 90	IN-1k : 20 : 384	187.0	88.0/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
UniRepLKNet-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	749.0	51.8	—	—	—	—	—
UniRepLKNet-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	835.0	53.0	—	—	—	—	—
UniRepLKNet-S	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	835.0	54.3	—	—	—	—	—
UniRepLKNet-B	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	978.0	54.8	—	—	—	—	—
UniRepLKNet-L	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1385.0	55.8	—	—	—	—	—
UniRepLKNet-XL	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1952.0	56.4	—	—	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
UniRepLKNet-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	749.0	44.9	—	—	—	—	—
UniRepLKNet-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	835.0	45.9	—	—	—	—	—
UniRepLKNet-S	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	835.0	47.1	—	—	—	—	—
UniRepLKNet-B	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	978.0	47.4	—	—	—	—	—
UniRepLKNet-L	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1385.0	48.4	—	—	—	—	—
UniRepLKNet-XL	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1952.0	49.0	—	—	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
UniRepLKNet-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	946.0	49.1	—	—	48.6	—	—
UniRepLKNet-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1036.0	51.0	—	—	50.5	—	—
UniRepLKNet-S	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 128 : 640	1618.0	52.7	—	—	51.9	—	—
UniRepLKNet-B	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 128 : 640	1850.0	53.9	—	—	53.5	—	—
UniRepLKNet-L	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 128 : 640	2507.0	55.1	—	—	54.5	—	—
UniRepLKNet-XL	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 128 : 640	3420.0	55.6	—	—	55.2	—	—