Heedless Backbones

Win 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

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
Win-T	28.5	IN-1k : Sup. : 300	— : — : —	4.56	82.3
Win-S	50.0	IN-1k : Sup. : 300	— : — : —	8.9	83.7
Win-B	88.0	IN-1k : Sup. : 300	— : — : —	15.6	83.5

model	params (m)	pretrain	finetune	gflops	IN-1k
Win-T	28.5	IN-1k : Sup. : 300	— : — : —	4.56	82.3/—
Win-S	50.0	IN-1k : Sup. : 300	— : — : —	8.9	83.7/—
Win-B	88.0	IN-1k : Sup. : 300	— : — : —	15.6	83.5/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
Win-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	268.0	46.6	68.4	51.1	—	—	—
Win-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	359.0	48.6	69.8	53.3	—	—	—
Win-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	503.0	49.1	70.2	53.8	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
Win-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	268.0	42.1	65.7	45.3	—	—	—
Win-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	359.0	43.4	67.3	46.7	—	—	—
Win-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	503.0	43.8	67.8	47.1	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
Win-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	949.0	47.2	—	—	45.7	—	—
Win-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1044.0	49.8	—	—	48.4	—	—
Win-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1196.0	50.7	—	—	49.2	—	—