Heedless Backbones

FlashSwin 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
FlashSwin-T-M8-p4	28.3	IN-1k : Sup. : 300	— : — : —	5.9	82.3
FlashSwin-T-M16-p4	28.3	IN-1k : Sup. : 300	— : — : —	6.6	82.9
FlashSwin-T-M32-p4	28.3	IN-1k : Sup. : 300	— : — : —	8.4	82.9
FlashSwin-T-M32-p2	28.3	IN-1k : Sup. : 300	— : — : —	37.4	84.1

model	params (m)	pretrain	finetune	gflops	IN-1k	IN-V2	IN-A	IN-R	IN-Sketch	IN-C↓
FlashSwin-T-M8-p4	28.3	IN-1k : Sup. : 300	— : — : —	5.9	82.3/—	70.8/—	28.1/—	43.6/—	30.4/—	56.9/—
FlashSwin-T-M16-p4	28.3	IN-1k : Sup. : 300	— : — : —	6.6	82.9/—	71.9/—	33.2/—	45.6/—	32.9/—	53.7/—
FlashSwin-T-M32-p4	28.3	IN-1k : Sup. : 300	— : — : —	8.4	82.9/—	72.0/—	32.9/—	45.6/—	32.4/—	53.0/—
FlashSwin-T-M32-p2	28.3	IN-1k : Sup. : 300	— : — : —	37.4	84.1/—	73.4/—	42.8/—	46.6/—	33.0/—	50.7/—
FlashSwin-L	197.0	IN-22k : Sup. : 90	IN-1k : — : 256	None	86.91/—	78.02/—	69.71/—	67.38/—	49.89/—	38.37/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
FlashSwin-T-M8-p4	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	38.5	59.8	41.4	18.1	42.9	57.2
FlashSwin-T-M16-p4	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	39.0	60.2	42.2	19.3	43.3	57.7
FlashSwin-T-M32-p4	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	40.3	61.7	43.6	20.1	44.7	59.7
FlashSwin-T-M32-p2	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	44.1	66.6	48.2	27.1	47.8	59.7

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
FlashSwin-T-M8-p4	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	35.3	56.8	37.6	12.8	38.2	57.2
FlashSwin-T-M16-p4	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	36.0	57.3	38.1	13.5	38.9	58.5
FlashSwin-T-M32-p4	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	36.7	58.6	39.3	14.4	39.9	59.7
FlashSwin-T-M32-p2	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	39.9	63.3	42.4	19.1	43.4	60.2

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
FlashSwin-T-M8-p4	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 256 : 512	None	—	—	—	44.39	—	—
FlashSwin-T-M16-p4	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 256 : 512	None	—	—	—	46.03	—	—
FlashSwin-T-M32-p4	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 256 : 512	None	—	—	—	46.88	—	—
FlashSwin-T-M32-p2	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 256 : 512	None	—	—	—	47.28	—	—
FlashSwin-L	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 256 : 512	None	—	—	—	52.58	—	—