Heedless Backbones

L2ViT 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
L2ViT-T	29.0	IN-1k : Sup. : 300	— : — : —	4.7	83.1
L2ViT-S	50.0	IN-1k : Sup. : 300	— : — : —	9.0	84.1
L2ViT-B	89.0	IN-1k : Sup. : 300	— : — : —	15.9	84.4
L2ViT-B	89.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	15.9	86.0
L2ViT-B	89.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	47.5	87.0

model	params (m)	pretrain	finetune	gflops	IN-1k
L2ViT-T	29.0	IN-1k : Sup. : 300	— : — : —	4.7	83.1/—
L2ViT-S	50.0	IN-1k : Sup. : 300	— : — : —	9.0	84.1/—
L2ViT-B	89.0	IN-1k : Sup. : 300	— : — : —	15.9	84.4/96.902
L2ViT-B	89.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	15.9	86.0/—
L2ViT-B	89.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	47.5	87.0/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
L2ViT-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	250.0	44.1	65.8	47.0	30.0	47.9	57.9
L2ViT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	269.0	45.5	68.6	50.1	—	—	—
L2ViT-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	341.0	46.2	68.0	49.6	31.7	50.0	61.1
L2ViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	360.0	47.2	70.0	51.7	—	—	—
L2ViT-B	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	484.0	46.5	68.7	49.4	31.2	50.6	60.7
L2ViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	504.0	47.5	70.5	51.8	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
L2ViT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	269.0	41.2	65.1	44.0	—	—	—
L2ViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	360.0	42.4	66.4	45.2	—	—	—
L2ViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	504.0	42.9	67.1	45.9	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
L2ViT-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	943.0	—	—	—	46.2	—	—
L2ViT-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1034.0	—	—	—	48.7	—	—
L2ViT-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1182.0	—	—	—	49.2	—	—