Heedless Backbones

VMamba 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

---------

Cityscapes (val)

Cityscapes (test)

ADE20K (val)

ADE20K (test)

metric

Select an option

mIoU^ms

pAcc^ms

mAcc^ms

mIoU^ss

pAcc^ss

mAcc^ss

GFLOPs

head

Select an option

---------

UPerNet

Mask2Former

Panoptic FPN

resolution

Select an option

----------

512x2048

640x2560

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

CL

Legend

Legend attribute

Select an option

Family

Pretrain Dataset

Pretrain Method

Semantic Segmentation Head

Semantic Segmentation Resolution

Semantic Segmentation Training Epochs

Legend attribute (second)

Select an option

----------

Family

Pretrain Method

Semantic Segmentation Head

Semantic Segmentation Resolution

Semantic Segmentation Training Epochs

model	params (m)	pretrain	head	train	GFLOPs	mIoU^ms
VMamba-T	30.0	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	949.0	48.8
VMamba-S	50.0	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1028.0	51.2
VMamba-B	89.0	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1170.0	51.6

model	params (m)	pretrain	finetune	gflops	IN-1k
VMamba-T	30.0	IN-1k : Sup. : 300	— : — : —	4.9	82.6/—
VMamba-S	50.0	IN-1k : Sup. : 300	— : — : —	8.7	83.6/—
VMamba-B	89.0	IN-1k : Sup. : 300	— : — : —	15.4	83.9/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
VMamba-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	271.0	47.3	69.3	52.0	—	—	—
VMamba-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	271.0	48.8	70.4	53.5	—	—	—
VMamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	349.0	48.7	70.0	53.4	—	—	—
VMamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	349.0	49.9	70.9	54.7	—	—	—
VMamba-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	485.0	49.2	71.4	54.0	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
VMamba-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	271.0	42.7	66.4	45.9	—	—	—
VMamba-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	271.0	43.7	67.4	47.0	—	—	—
VMamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	349.0	43.7	67.3	47.0	—	—	—
VMamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	349.0	44.2	68.2	47.7	—	—	—
VMamba-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	485.0	44.1	68.3	47.7	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
VMamba-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	949.0	48.8	—	—	47.9	—	—
VMamba-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1028.0	51.2	—	—	50.6	—	—
VMamba-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1170.0	51.6	—	—	51.0	—	—