Heedless Backbones

PlainMamba 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

---------

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

Classification Resolution

Legend attribute (second)

Select an option

----------

Family

Pretrain Method

Classification Resolution

model	params (m)	pretrain	finetune	GFLOPs	Top-1
PlainMamba-L1	7.3	IN-1k : Sup. : 300	— : — : —	3.0	77.9
PlainMamba-L2	25.7	IN-1k : Sup. : 300	— : — : —	8.1	81.6
PlainMamba-L3	50.5	IN-1k : Sup. : 300	— : — : —	14.4	82.3

model	params (m)	pretrain	finetune	gflops	IN-1k
PlainMamba-L1	7.3	IN-1k : Sup. : 300	— : — : —	3.0	77.9/—
PlainMamba-L2	25.7	IN-1k : Sup. : 300	— : — : —	8.1	81.6/—
PlainMamba-L3	50.5	IN-1k : Sup. : 300	— : — : —	14.4	82.3/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
PlainMamba-L1	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	388.0	44.1	64.8	47.9	—	—	—
PlainMamba-L1	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	250.0	41.7	62.1	44.4	—	—	—
PlainMamba-L2	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	542.0	46.0	66.9	50.1	—	—	—
PlainMamba-L2	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	392.0	43.9	64.9	47.0	—	—	—
PlainMamba-L3	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	696.0	46.8	68.0	51.1	—	—	—
PlainMamba-L3	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	478.0	44.8	66.0	47.9	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
PlainMamba-L1	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	388.0	39.1	61.6	41.9	—	—	—
PlainMamba-L2	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	542.0	40.6	63.8	43.6	—	—	—
PlainMamba-L3	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	696.0	41.2	64.7	43.9	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
PlainMamba-L1	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	174.0	—	—	—	44.1	—	—
PlainMamba-L2	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	285.0	—	—	—	46.8	—	—
PlainMamba-L3	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	419.0	—	—	—	49.1	—	—