Heedless Backbones

FMViT 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

Classification Resolution

Legend attribute (second)

Select an option

----------

Family

Pretrain Method

Classification Resolution

model	params (m)	pretrain	finetune	GFLOPs	Top-1
FMViT-T	2.0	IN-1k : Sup. : 300	— : — : —	0.3	72.9
FMViT-S	6.4	IN-1k : Sup. : 300	— : — : —	0.8	78.5
FMViT-M	12.8	IN-1k : Sup. : 300	— : — : —	2.0	80.3
FMViT-B	24.3	IN-1k : Sup. : 300	— : — : —	4.2	82.4
FMViT-L	35.3	IN-1k : Sup. : 300	— : — : —	7.1	83.3

model	params (m)	pretrain	finetune	gflops	IN-1k
FMViT-T	2.0	IN-1k : Sup. : 300	— : — : —	0.3	72.9/—
FMViT-S	6.4	IN-1k : Sup. : 300	— : — : —	0.8	78.5/—
FMViT-M	12.8	IN-1k : Sup. : 300	— : — : —	2.0	80.3/—
FMViT-B	24.3	IN-1k : Sup. : 300	— : — : —	4.2	82.4/—
FMViT-L	35.3	IN-1k : Sup. : 300	— : — : —	7.1	83.3/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
FMViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	269.9	44.1	66.1	48.0	—	—	—
FMViT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	333.0	46.4	68.1	51.2	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
FMViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	269.9	41.8	65.1	45.0	—	—	—
FMViT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	333.0	41.9	65.4	45.1	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
FMViT-B	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	22.2	—	—	—	43.5	—	—
FMViT-L	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	37.4	—	—	—	46.9	—	—