Heedless Backbones

LaViT 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
LaViT-T	10.9	IN-1k : Sup. : 300	— : — : —	1.6	79.2
LaViT-S	22.4	IN-1k : Sup. : 300	— : — : —	3.3	82.6
LaViT-B	39.6	IN-1k : Sup. : 300	— : — : —	6.1	83.1

model	params (m)	pretrain	finetune	gflops	IN-1k
LaViT-T	10.9	IN-1k : Sup. : 300	— : — : —	1.6	79.2/—
LaViT-S	22.4	IN-1k : Sup. : 300	— : — : —	3.3	82.6/—
LaViT-B	39.6	IN-1k : Sup. : 300	— : — : —	6.1	83.1/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
LaViT-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	202.0	46.2	67.2	49.1	29.6	50.2	61.3
LaViT-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 36	202.0	48.4	69.9	51.7	31.8	52.2	64.1
LaViT-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	290.0	46.7	68.3	49.7	29.9	50.7	61.7
LaViT-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 36	290.0	48.9	70.3	52.2	33.1	52.6	65.4

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
LaViT-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	920.0	49.5	—	—	47.2	—	—
LaViT-S	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	122.0	—	—	—	44.1	—	—