Heedless Backbones

iFormer 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
iFormer-S	20.0	IN-1k : Sup. : 300	— : — : —	4.8	83.4
iFormer-S	20.0	IN-1k : Sup. : 300	IN-1k : — : 384	16.1	84.6
iFormer-B	48.0	IN-1k : Sup. : 300	— : — : —	9.4	84.6
iFormer-B	48.0	IN-1k : Sup. : 300	IN-1k : — : 384	30.5	85.7
iFormer-L	87.0	IN-1k : Sup. : 300	— : — : —	14.0	84.8
iFormer-L	87.0	IN-1k : Sup. : 300	IN-1k : — : 384	45.3	85.8

model	params (m)	pretrain	finetune	gflops	IN-1k
iFormer-S	20.0	IN-1k : Sup. : 300	— : — : —	4.8	83.4/96.6
iFormer-S	20.0	IN-1k : Sup. : 300	IN-1k : — : 384	16.1	84.6/—
iFormer-B	48.0	IN-1k : Sup. : 300	— : — : —	9.4	84.6/97.0
iFormer-B	48.0	IN-1k : Sup. : 300	IN-1k : — : 384	30.5	85.7/—
iFormer-L	87.0	IN-1k : Sup. : 300	— : — : —	14.0	84.8/97.0
iFormer-L	87.0	IN-1k : Sup. : 300	IN-1k : — : 384	45.3	85.8/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
iFormer-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	263.0	46.2	68.5	50.6	—	—	—
iFormer-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	351.0	48.3	70.3	53.2	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
iFormer-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	263.0	41.9	65.3	45.0	—	—	—
iFormer-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	351.0	43.4	67.2	46.7	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
iFormer-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	938.0	48.8	—	—	48.4	—	—
iFormer-S	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	181.0	—	—	—	48.6	—	—