Heedless Backbones

EfficientVMamba 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)

PASCAL VOC 2007 (val)

PASCAL VOC 2007 (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

SETR

resolution

Select an option

----------

512x2048

640x2560

axis

Select an option

Results

Parameters (M)

Images / Second

GFLOPs

Publication Date

Filters

pretrain dataset

Select an option

---------

MegData73M

JFT-3B

JFT-300M

ImageNet-1k

ImageNet-22k

pretrain method

Select an option

----------

Supervised

Sup. + TL

FCMAE

MAE

CL

Legend

Legend attribute

Select an option

Family

Pretrain Dataset

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
EfficientVMamba-T	6.0	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	230.0	39.3
EfficientVMamba-S	11.0	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	505.0	42.1
EfficientVMamba-B	33.0	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	930.0	47.3

model	params (m)	pretrain	finetune	gflops	IN-1k
EfficientVMamba-T	6.0	IN-1k : Sup. : 300	— : — : —	0.8	76.5/—
EfficientVMamba-S	11.0	IN-1k : Sup. : 300	— : — : —	1.3	78.7/—
EfficientVMamba-B	33.0	IN-1k : Sup. : 300	— : — : —	4.0	81.8/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
EfficientVMamba-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	37.5	57.8	39.6	22.6	40.7	49.1
EfficientVMamba-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	60.0	35.6	57.7	38.0	—	—	—
EfficientVMamba-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	60.0	38.3	60.3	41.6	—	—	—
EfficientVMamba-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	39.1	60.3	41.2	23.9	43.0	51.3
EfficientVMamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	197.0	39.3	61.8	42.6	—	—	—
EfficientVMamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	197.0	41.6	63.9	45.6	—	—	—
EfficientVMamba-B	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	42.8	63.9	45.8	27.3	46.9	55.1
EfficientVMamba-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	252.0	43.7	66.2	47.9	—	—	—
EfficientVMamba-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	252.0	45.0	66.9	49.2	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
EfficientVMamba-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	60.0	33.2	54.4	35.1	—	—	—
EfficientVMamba-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	60.0	35.3	57.2	37.6	—	—	—
EfficientVMamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	197.0	36.7	58.9	39.2	—	—	—
EfficientVMamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	197.0	38.2	60.8	40.7	—	—	—
EfficientVMamba-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	252.0	40.2	63.3	42.9	—	—	—
EfficientVMamba-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	252.0	40.8	64.1	43.7	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
EfficientVMamba-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	230.0	39.3	—	—	38.9	—	—
EfficientVMamba-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	505.0	42.1	—	—	41.5	—	—
EfficientVMamba-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	930.0	47.3	—	—	46.5	—	—