Iwin Transformer 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

---------

Cityscapes (val)

Cityscapes (test)

ADE20K (val)

ADE20K (test)

PASCAL VOC 2007 (val)

PASCAL VOC 2007 (test)

metric

Select an option

mIoU^ms

pAcc^ms

mAcc^ms

mIoU^ss

pAcc^ss

mAcc^ss

GFLOPs

head

Select an option

---------

UPerNet

Mask2Former

Panoptic FPN

SETR

resolution

Select an option

----------

512x2048

640x2560

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

MAP

Legend

Legend attribute

Select an option

Family

Pretrain Dataset

Semantic Segmentation Head

Semantic Segmentation Resolution

Semantic Segmentation Training Epochs

Legend attribute (second)

Select an option

----------

Family

Pretrain Method

Semantic Segmentation Head

Semantic Segmentation Resolution

Semantic Segmentation Training Epochs

No Results

model	params (m)	pretrain	finetune	gflops	IN-1k
Iwin-T	30.2	IN-1k : Sup. : 300	— : — : —	4.7	82.0/—
Iwin-S	51.6	IN-1k : Sup. : 300	— : — : —	9.0	83.4/—
Iwin-S	51.6	IN-1k : Sup. : 300	IN-1k : 30 : 384	27.7	84.3/—
Iwin-S	51.6	IN-1k : Sup. : 300	IN-1k : 30 : 512	52.0	84.4/—
Iwin-S	51.6	IN-1k : Sup. : 300	IN-1k : 30 : 1024	207.9	83.8/—
Iwin-B	91.2	IN-1k : Sup. : 300	— : — : —	15.9	83.5/—
Iwin-B	91.2	IN-1k : Sup. : 300	IN-1k : 30 : 384	48.3	84.9/—
Iwin-B	91.2	IN-1k : Sup. : 300	IN-1k : 30 : 512	89.5	85.1/—
Iwin-B	91.2	IN-1k : Sup. : 300	IN-1k : 30 : 1024	358.2	85.0/—
Iwin-B	91.2	IN-22k : Sup. : 90	IN-1k : 30 : 224	15.9	85.5/—
Iwin-B	91.2	IN-22k : Sup. : 90	IN-1k : 30 : 384	48.3	86.6/—
Iwin-B	91.2	IN-22k : Sup. : 90	IN-1k : 30 : 512	89.5	86.1/—
Iwin-B	91.2	IN-22k : Sup. : 90	IN-1k : 30 : 1024	358.2	85.6/—
Iwin-L	204.3	IN-22k : Sup. : 90	IN-1k : 30 : 224	35.4	86.4/—
Iwin-L	204.3	IN-22k : Sup. : 90	IN-1k : 30 : 384	106.6	87.4/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
Iwin-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	268.0	42.2	65.3	45.8	—	—	—
Iwin-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	268.0	44.7	67.2	48.8	—	—	—
Iwin-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 12	747.0	47.2	66.1	51.3	—	—	—
Iwin-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	747.0	49.4	68.4	53.5	—	—	—
Iwin-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	358.0	43.7	67.0	47.4	—	—	—
Iwin-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	358.0	45.5	67.5	49.6	—	—	—
Iwin-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	837.0	49.4	68.1	53.3	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
Iwin-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	268.0	38.9	62.1	41.6	—	—	—
Iwin-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	268.0	40.9	64.1	43.6	—	—	—
Iwin-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 12	747.0	40.9	63.5	44.1	—	—	—
Iwin-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	747.0	42.9	65.8	46.4	—	—	—
Iwin-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	358.0	40.0	63.9	42.5	—	—	—
Iwin-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	358.0	41.0	64.3	44.0	—	—	—
Iwin-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	837.0	43.0	65.6	46.4	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
Iwin-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	946.0	—	—	—	44.7	—	—
Iwin-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1038.0	—	—	—	47.5	—	—
Iwin-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1189.0	—	—	—	48.9	—	—