Heedless Backbones

LightViT 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
LightViT-T	9.4	IN-1k : Sup. : 300	— : — : —	0.7	78.7
LightViT-S	19.2	IN-1k : Sup. : 300	— : — : —	1.7	80.8
LightViT-B	35.2	IN-1k : Sup. : 300	— : — : —	3.9	82.1

model	params (m)	pretrain	finetune	gflops	IN-1k
LightViT-T	9.4	IN-1k : Sup. : 300	— : — : —	0.7	78.7/94.4
LightViT-S	19.2	IN-1k : Sup. : 300	— : — : —	1.7	80.8/95.3
LightViT-B	35.2	IN-1k : Sup. : 300	— : — : —	3.9	82.1/95.9

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
LightViT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	187.0	37.8	60.7	40.4	—	—	—
LightViT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	187.0	41.5	64.4	45.1	—	—	—
LightViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	204.0	40.0	62.9	42.6	—	—	—
LightViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	204.0	43.2	66.0	47.4	—	—	—
LightViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	240.0	41.7	64.5	45.1	—	—	—
LightViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	240.0	45.0	67.9	48.8	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
LightViT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	187.0	35.9	57.8	38.0	—	—	—
LightViT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	187.0	38.4	61.2	40.8	—	—	—
LightViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	204.0	37.4	60.0	39.3	—	—	—
LightViT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	204.0	39.9	63.0	42.7	—	—	—
LightViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	240.0	38.8	61.4	41.4	—	—	—
LightViT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	240.0	41.2	64.8	44.2	—	—	—