Heedless Backbones

SMT 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

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
SMT-M	6.5	IN-1k : Sup. : 300	— : — : —	1.3	78.4
SMT-T	11.5	IN-1k : Sup. : 300	— : — : —	2.4	82.2
SMT-S	20.5	IN-1k : Sup. : 300	— : — : —	4.7	83.7
SMT-B	32.0	IN-1k : Sup. : 300	— : — : —	7.7	84.3
SMT-L	80.5	IN-1k : Sup. : 300	— : — : —	17.7	84.6
SMT-L	80.5	IN-22k : Sup. : 90	IN-1k : 30 : 224	17.7	87.1
SMT-L	80.5	IN-22k : Sup. : 90	IN-1k : 30 : 384	54.6	88.1

model	params (m)	pretrain	finetune	gflops	IN-1k
SMT-M	6.5	IN-1k : Sup. : 300	— : — : —	1.3	78.4/—
SMT-T	11.5	IN-1k : Sup. : 300	— : — : —	2.4	82.2/96.0
SMT-S	20.5	IN-1k : Sup. : 300	— : — : —	4.7	83.7/96.5
SMT-B	32.0	IN-1k : Sup. : 300	— : — : —	7.7	84.3/96.9
SMT-L	80.5	IN-1k : Sup. : 300	— : — : —	17.7	84.6/—
SMT-L	80.5	IN-22k : Sup. : 90	IN-1k : 30 : 224	17.7	87.1/98.1
SMT-L	80.5	IN-22k : Sup. : 90	IN-1k : 30 : 384	54.6	88.1/98.4

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
SMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	265.0	47.8	69.5	52.1	—	—	—
SMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	265.0	49.0	70.1	53.4	—	—	—
SMT-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	744.0	51.9	70.5	56.3	—	—	—
SMT-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 36	247.0	47.3	67.8	50.5	32.5	51.1	62.3
SMT-S	IN-1k : Sup. : 300	Sparse R-CNN	COCO (train) : 36	171.0	50.2	69.8	54.7	—	—	—
SMT-S	IN-1k : Sup. : 300	ATSS	COCO (train) : 36	214.0	49.9	68.9	54.7	—	—	—
SMT-S	IN-1k : Sup. : 300	DINO	COCO (train) : 12	309.0	54.0	71.9	59.0	—	—	—
SMT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	328.0	49.0	70.2	53.7	—	—	—
SMT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	328.0	49.8	71.0	54.4	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
SMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	265.0	43.0	66.6	46.1	—	—	—
SMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	265.0	43.4	67.3	46.7	—	—	—
SMT-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	744.0	44.7	67.8	48.6	—	—	—
SMT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	328.0	44.0	67.6	47.4	—	—	—
SMT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	328.0	44.0	68.0	47.3	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
SMT-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	935.0	50.2	—	—	49.2	—	—
SMT-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1004.0	50.6	—	—	49.6	—	—