Heedless Backbones

A2Mamba 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

---------

ImageNet-1k

ImageNet-A

ImageNet-R

ImageNet-Sketch

ImageNet-C

ImageNet-C-bar

ImageNet-V2

ImageNet-ReaL

PASCAL VOC 2007 (val)

PASCAL VOC 2007 (test)

metric

Select an option

----------

Top-1

Top-5

GFLOPs

resolution

Select an option

----------

224x224

384x384

512x512

axis

Select an option

Results

Parameters (M)

Images / Second

GFLOPs

Publication Date

Filters

pretrain dataset

Select an option

---------

ImageNet-1k

ImageNet-22k

JFT-300M

JFT-3B

MegData73M

pretrain method

Select an option

----------

Supervised

Sup. + TL

FCMAE

MAE

CL

MAP

Legend

Legend attribute

Select an option

Family

Pretrain Dataset

Pretrain Method

Classification Resolution

Legend attribute (second)

Select an option

----------

Pretrain Dataset

Pretrain Method

Classification Resolution

model	params (m)	pretrain	finetune	GFLOPs	Top-1
A2Mamba-N	4.0	IN-1k : Sup. : 300	— : — : —	0.8	78.7
A2Mamba-T	15.0	IN-1k : Sup. : 300	— : — : —	2.7	83.0
A2Mamba-S	31.0	IN-1k : Sup. : 300	— : — : —	5.4	84.7
A2Mamba-B	51.0	IN-1k : Sup. : 300	— : — : —	10.7	85.7
A2Mamba-B	51.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	34.0	86.4
A2Mamba-L	95.0	IN-1k : Sup. : 300	— : — : —	17.4	86.1
A2Mamba-L	95.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	54.0	86.7

model	params (m)	pretrain	finetune	gflops	IN-1k
A2Mamba-N	4.0	IN-1k : Sup. : 300	— : — : —	0.8	78.7/—
A2Mamba-T	15.0	IN-1k : Sup. : 300	— : — : —	2.7	83.0/—
A2Mamba-S	31.0	IN-1k : Sup. : 300	— : — : —	5.4	84.7/—
A2Mamba-B	51.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	34.0	86.4/—
A2Mamba-B	51.0	IN-1k : Sup. : 300	— : — : —	10.7	85.7/—
A2Mamba-L	95.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	54.0	86.7/—
A2Mamba-L	95.0	IN-1k : Sup. : 300	— : — : —	17.4	86.1/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
A2Mamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	283.0	51.5	—	—	—	—	—
A2Mamba-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	762.0	54.0	—	—	—	—	—
A2Mamba-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	410.0	52.7	—	—	—	—	—
A2Mamba-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	889.0	55.4	—	—	—	—	—
A2Mamba-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	552.0	53.0	—	—	—	—	—
A2Mamba-L	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	1027.0	55.6	—	—	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
A2Mamba-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	283.0	45.3	—	—	—	—	—
A2Mamba-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	762.0	46.6	—	—	—	—	—
A2Mamba-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	410.0	46.8	—	—	—	—	—
A2Mamba-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	889.0	47.6	—	—	—	—	—
A2Mamba-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	552.0	46.8	—	—	—	—	—
A2Mamba-L	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	1027.0	48.1	—	—	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
A2Mamba-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	959.0	52.0	—	—	51.6	—	—
A2Mamba-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1090.0	53.9	—	—	53.3	—	—
A2Mamba-L	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1237.0	54.1	—	—	53.7	—	—