SSViT 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

MAP

Legend

Legend attribute

Select an option

Family

Pretrain Dataset

Pretrain Method

Classification Resolution

Legend attribute (second)

Select an option

----------

Pretrain Dataset

Pretrain Method

Classification Resolution

model	params (m)	pretrain	finetune	GFLOPs	Top-1
SSViT-T	15.0	IN-1k : Sup. : 300	— : — : —	2.4	83.0
SSViT-S	27.0	IN-1k : Sup. : 300	— : — : —	4.4	84.4
SSViT-B	57.0	IN-1k : Sup. : 300	— : — : —	9.6	85.3
SSViT-L	100.0	IN-1k : Sup. : 300	— : — : —	18.2	85.7

model	params (m)	pretrain	finetune	gflops	IN-1k	IN-V2	IN-A	IN-R
SSViT-T	15.0	IN-1k : Sup. : 300	— : — : —	2.4	83.0/—	72.3/—	32.6/—	45.6/—
SSViT-S	27.0	IN-1k : Sup. : 300	— : — : —	4.4	84.4/—	74.1/—	41.6/—	51.0/—
SSViT-B	57.0	IN-1k : Sup. : 300	— : — : —	9.6	85.3/—	75.7/—	49.4/—	55.6/—
SSViT-L	100.0	IN-1k : Sup. : 300	— : — : —	18.2	85.7/—	76.1/—	55.0/—	59.2/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
SSViT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	223.0	47.3	69.1	51.7	—	—	—
SSViT-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	205.0	45.6	66.5	49.3	28.6	50.1	60.5
SSViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	266.0	51.2	72.0	56.0	—	—	—
SSViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	266.0	49.4	70.8	54.1	—	—	—
SSViT-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	745.0	53.8	72.4	58.1	—	—	—
SSViT-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	248.0	47.5	68.6	50.8	30.1	52.2	63.3
SSViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	382.0	52.6	73.2	57.7	—	—	—
SSViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	382.0	51.0	72.5	55.8	—	—	—
SSViT-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	861.0	54.9	73.7	59.7	—	—	—
SSViT-B	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	363.0	49.0	70.2	52.9	32.4	53.4	64.8
SSViT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	572.0	51.6	72.9	56.6	—	—	—
SSViT-L	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	553.0	50.0	71.4	53.8	33.2	54.6	65.0

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
SSViT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	223.0	42.6	66.2	45.8	—	—	—
SSViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	266.0	45.4	69.7	49.0	—	—	—
SSViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	266.0	44.0	67.7	47.3	—	—	—
SSViT-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	745.0	46.6	70.1	50.4	—	—	—
SSViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	382.0	46.4	70.9	50.3	—	—	—
SSViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	382.0	45.4	69.7	48.9	—	—	—
SSViT-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	861.0	47.6	71.6	51.5	—	—	—
SSViT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	572.0	46.0	70.1	49.8	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
SSViT-T	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 640 : 512	35.0	—	—	—	46.8	—	—
SSViT-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 160 : 512	941.0	50.1	—	—	—	—	—
SSViT-S	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 640 : 512	184.0	—	—	—	49.6	—	—
SSViT-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 160 : 512	1060.0	52.2	—	—	—	—	—
SSViT-B	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 640 : 512	303.0	—	—	—	51.0	—	—
SSViT-L	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 160 : 512	1256.0	53.3	—	—	—	—	—
SSViT-L	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 640 : 512	497.0	—	—	—	51.5	—	—