Heedless Backbones

LSNet 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
LSNet-T	11.4	IN-1k : Sup. : 300	— : — : —	0.3	74.9
LSNet-S	16.1	IN-1k : Sup. : 300	— : — : —	0.5	77.8
LSNet-B	23.2	IN-1k : Sup. : 300	— : — : —	1.3	80.3
LSNet-T	11.4	IN-1k : Sup. : 300	— : — : —	0.3	76.1
LSNet-S	16.1	IN-1k : Sup. : 300	— : — : —	0.5	79.0
LSNet-B	23.2	IN-1k : Sup. : 300	— : — : —	1.3	81.6

model	params (m)	pretrain	finetune	gflops	IN-1k	IN-C↓	IN-A	IN-R	IN-Sketch
LSNet-T	11.4	IN-1k : Sup. : 300	— : — : —	0.3	74.9/—	68.2/—	6.7/—	38.5/—	25.5/—
LSNet-S	16.1	IN-1k : Sup. : 300	— : — : —	0.5	77.8/—	65.7/—	9.6/—	39.4/—	27.5/—
LSNet-B	23.2	IN-1k : Sup. : 300	— : — : —	1.3	80.3/—	59.3/—	17.3/—	43.1/—	30.7/—
LSNet-T	11.4	IN-1k : Sup. : 300	— : — : —	0.3	76.1/—	—/—	—/—	—/—	—/—
LSNet-S	16.1	IN-1k : Sup. : 300	— : — : —	0.5	79.0/—	—/—	—/—	—/—	—/—
LSNet-B	23.2	IN-1k : Sup. : 300	— : — : —	1.3	81.6/—	—/—	—/—	—/—	—/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
LSNet-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	34.2	54.6	35.2	17.8	37.1	48.5
LSNet-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	35.0	57.0	37.3	—	—	—
LSNet-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	36.7	57.2	38.6	20.0	39.7	51.8
LSNet-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	37.4	59.9	39.8	—	—	—
LSNet-B	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	39.2	60.0	41.5	22.1	43.0	52.9
LSNet-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	40.8	63.4	44.0	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
LSNet-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	32.7	53.8	34.3	—	—	—
LSNet-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	34.8	56.8	36.6	—	—	—
LSNet-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	37.8	60.5	40.1	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
LSNet-T	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	None	—	—	—	40.1	—	—
LSNet-S	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	None	—	—	—	41.0	—	—
LSNet-B	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	None	—	—	—	43.0	—	—