Heedless Backbones

MVFormer 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
MVFormer-T	27.0	IN-1k : Sup. : 300	— : — : —	3.9	83.4
MVFormer-T	27.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	11.5	84.8
MVFormer-S	40.0	IN-1k : Sup. : 300	— : — : —	7.6	84.3
MVFormer-S	40.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	22.2	85.6
MVFormer-B	57.0	IN-1k : Sup. : 300	— : — : —	12.7	84.6
MVFormer-B	57.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	37.4	85.7

model	params (m)	pretrain	finetune	gflops	IN-1k
MVFormer-T	27.0	IN-1k : Sup. : 300	— : — : —	3.9	83.4/—
MVFormer-T	27.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	11.5	84.8/—
MVFormer-S	40.0	IN-1k : Sup. : 300	— : — : —	7.6	84.3/—
MVFormer-S	40.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	22.2	85.6/—
MVFormer-B	57.0	IN-1k : Sup. : 300	— : — : —	12.7	84.6/—
MVFormer-B	57.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	37.4	85.7/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
MVFormer-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	231.0	46.2	68.4	50.8	—	—	—
MVFormer-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	231.0	44.8	66.1	48.0	—	—	—
MVFormer-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	231.0	47.6	69.1	52.0	—	—	—
MVFormer-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	299.0	47.5	69.8	52.3	—	—	—
MVFormer-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	306.0	45.8	67.2	49.3	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
MVFormer-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	231.0	42.1	65.4	45.2	—	—	—
MVFormer-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	231.0	43.1	66.2	46.5	—	—	—
MVFormer-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	299.0	43.0	66.8	46.3	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
MVFormer-T	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	42.0	—	—	—	47.1	—	—
MVFormer-S	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	61.0	—	—	—	48.8	—	—