Heedless Backbones

iFormer (Mobile) 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-T (Mobile)	2.9	IN-1k : Sup. : 300	— : — : —	0.53	74.1
iFormer-S (Mobile)	6.5	IN-1k : Sup. : 300	— : — : —	1.09	78.8
iFormer-M (Mobile)	8.9	IN-1k : Sup. : 300	— : — : —	1.64	80.4
iFormer-L (Mobile)	14.7	IN-1k : Sup. : 300	— : — : —	2.63	81.9
iFormer-M (Mobile)	8.9	IN-1k : Sup. : 300	— : — : —	1.64	81.1
iFormer-L (Mobile)	14.7	IN-1k : Sup. : 300	— : — : —	2.63	82.7
iFormer-L2 (Mobile)	24.5	IN-1k : Sup. : 450	— : — : —	4.5	83.9
iFormer-H (Mobile)	99.0	IN-1k : Sup. : 300	— : — : —	15.5	84.8

model	params (m)	pretrain	finetune	gflops	IN-1k
iFormer-T (Mobile)	2.9	IN-1k : Sup. : 300	— : — : —	0.53	74.1/—
iFormer-S (Mobile)	6.5	IN-1k : Sup. : 300	— : — : —	1.09	78.8/—
iFormer-M (Mobile)	8.9	IN-1k : Sup. : 300	— : — : —	1.64	80.4/95.336
iFormer-L (Mobile)	14.7	IN-1k : Sup. : 300	— : — : —	2.63	81.9/—
iFormer-M (Mobile)	8.9	IN-1k : Sup. : 300	— : — : —	1.64	81.1/95.466
iFormer-L (Mobile)	14.7	IN-1k : Sup. : 300	— : — : —	2.63	82.7/—
iFormer-L2 (Mobile)	24.5	IN-1k : Sup. : 450	— : — : —	4.5	83.9/—
iFormer-H (Mobile)	99.0	IN-1k : Sup. : 300	— : — : —	15.5	84.8/97.058

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
iFormer-M (Mobile)	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	40.8	62.5	44.8	—	—	—
iFormer-L (Mobile)	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	42.2	64.2	46.0	—	—	—
iFormer-L2 (Mobile)	IN-1k : Sup. : 450	Mask R-CNN	COCO (train) : 12	None	44.6	66.7	49.1	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
iFormer-M (Mobile)	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	37.9	59.7	40.7	—	—	—
iFormer-L (Mobile)	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	39.1	61.4	41.9	—	—	—
iFormer-L2 (Mobile)	IN-1k : Sup. : 450	Mask R-CNN	COCO (train) : 12	None	41.1	64.0	44.1	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
iFormer-M (Mobile)	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	None	—	—	—	42.4	—	—
iFormer-L (Mobile)	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	None	—	—	—	44.5	—	—
iFormer-L2 (Mobile)	IN-1k : Sup. : 450	Panoptic FPN	ADE20K (train) : 64 : 512	None	—	—	—	46.2	—	—