Heedless Backbones

FLatten-Swin 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
FLatten-Swin-T	29.0	IN-1k : Sup. : 300	— : — : —	4.5	82.1
FLatten-Swin-S	51.0	IN-1k : Sup. : 300	— : — : —	8.7	83.5
FLatten-Swin-B	89.0	IN-1k : Sup. : 300	— : — : —	15.4	83.8
FLatten-Swin-B	89.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	46.5	85.0

model	params (m)	pretrain	finetune	gflops	IN-1k
FLatten-Swin-T	29.0	IN-1k : Sup. : 300	— : — : —	4.5	82.1/95.9
FLatten-Swin-S	51.0	IN-1k : Sup. : 300	— : — : —	8.7	83.5/—
FLatten-Swin-B	89.0	IN-1k : Sup. : 300	— : — : —	15.4	83.8/—
FLatten-Swin-B	89.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	46.5	85.0/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
FLatten-Swin-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	268.0	44.2	67.3	48.5	29.4	47.5	57.0
FLatten-Swin-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	268.0	46.5	68.5	50.8	31.2	49.6	60.4
FLatten-Swin-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	747.0	50.8	69.6	55.1	34.2	54.6	65.5
FLatten-Swin-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	841.0	52.2	71.2	56.8	35.6	56.4	67.6

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
FLatten-Swin-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	268.0	40.2	63.8	43.0	24.5	43.8	54.7
FLatten-Swin-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	268.0	42.1	65.4	45.1	25.4	45.4	56.8
FLatten-Swin-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	747.0	44.1	67.0	48.1	27.6	48.1	59.0
FLatten-Swin-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	841.0	45.4	68.3	49.4	29.3	49.0	60.8

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
FLatten-Swin-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	946.0	—	—	—	44.82	—	57.01
FLatten-Swin-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1038.0	—	—	—	48.14	—	59.31