Heedless Backbones

STViT 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
STViT-S	25.4	IN-1k : Sup. : 300	— : — : —	4.37	83.6
STViT-S	25.4	IN-1k : Sup. : 300	IN-1k : 30 : 384	14.1	85.0
STViT-B	52.0	IN-1k : Sup. : 300	— : — : —	9.9	84.8
STViT-B	52.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	31.5	86.0
STViT-L	95.0	IN-1k : Sup. : 300	— : — : —	15.6	85.3
STViT-L	95.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	49.7	86.4

model	params (m)	pretrain	finetune	gflops	IN-1k
STViT-S	25.4	IN-1k : Sup. : 300	— : — : —	4.37	83.6/—
STViT-S	25.4	IN-1k : Sup. : 300	IN-1k : 30 : 384	14.1	85.0/—
STViT-B	52.0	IN-1k : Sup. : 300	— : — : —	9.9	84.8/—
STViT-B	52.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	31.5	86.0/—
STViT-L	95.0	IN-1k : Sup. : 300	— : — : —	15.6	85.3/—
STViT-L	95.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	49.7	86.4/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
STViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	252.0	47.6	70.0	52.3	—	—	—
STViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	252.0	49.2	70.8	54.4	—	—	—
STViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	359.0	49.7	71.7	54.7	—	—	—
STViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	359.0	51.0	72.3	56.0	—	—	—
STViT-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	837.0	53.9	72.7	58.5	—	—	—
STViT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	470.0	50.8	72.5	56.3	—	—	—
STViT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	470.0	51.7	73.0	56.9	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
STViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	252.0	43.1	66.8	46.5	—	—	—
STViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	252.0	44.2	68.0	47.7	—	—	—
STViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	359.0	44.8	68.9	48.7	—	—	—
STViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	359.0	45.4	69.5	49.3	—	—	—
STViT-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	837.0	46.8	70.4	50.8	—	—	—
STViT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	470.0	45.5	69.7	49.1	—	—	—
STViT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	470.0	45.9	70.4	49.9	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
STViT-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	926.0	49.0	—	—	48.6	—	—
STViT-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1036.0	51.9	—	—	50.7	—	—
STViT-L	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1151.0	53.2	—	—	52.4	—	—