RMT 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

---------

MegData73M

JFT-3B

JFT-300M

ImageNet-1k

ImageNet-22k

pretrain method

Select an option

----------

Supervised

Sup. + TL

FCMAE

MAE

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
RMT-T	14.0	IN-1k : Sup. : 300	— : — : —	2.5	82.4
RMT-S	27.0	IN-1k : Sup. : 300	— : — : —	4.5	84.1
RMT-S	27.0	IN-1k : Sup. + TL : 300	— : — : —	4.5	84.8
RMT-B	54.0	IN-1k : Sup. : 300	— : — : —	9.7	85.0
RMT-B	54.0	IN-1k : Sup. + TL : 300	— : — : —	9.7	85.6
RMT-L	95.0	IN-1k : Sup. : 300	— : — : —	18.2	85.5
RMT-L	95.0	IN-1k : Sup. + TL : 300	— : — : —	18.2	86.1

model	params (m)	pretrain	finetune	gflops	IN-1k
RMT-T	14.0	IN-1k : Sup. : 300	— : — : —	2.5	82.4/—
RMT-S	27.0	IN-1k : Sup. : 300	— : — : —	4.5	84.1/—
RMT-S	27.0	IN-1k : Sup. + TL : 300	— : — : —	4.5	84.8/—
RMT-B	54.0	IN-1k : Sup. : 300	— : — : —	9.7	85.0/—
RMT-B	54.0	IN-1k : Sup. + TL : 300	— : — : —	9.7	85.6/—
RMT-L	95.0	IN-1k : Sup. : 300	— : — : —	18.2	85.5/—
RMT-L	95.0	IN-1k : Sup. + TL : 300	— : — : —	18.2	86.1/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
RMT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	218.0	47.1	68.8	51.7	—	—	—
RMT-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	199.0	45.1	66.2	48.1	28.8	48.9	61.1
RMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	262.0	49.0	70.8	53.9	—	—	—
RMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	262.0	50.7	71.9	55.6	—	—	—
RMT-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	244.0	47.8	69.1	51.8	32.1	51.8	63.5
RMT-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	741.0	53.2	72.0	57.8	—	—	—
RMT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	373.0	51.1	72.5	56.1	—	—	—
RMT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	373.0	52.2	72.9	57.0	—	—	—
RMT-B	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	355.0	49.1	70.3	53.0	32.9	53.2	64.2
RMT-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	852.0	54.5	72.8	59.0	—	—	—
RMT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	557.0	51.6	73.1	56.5	—	—	—
RMT-L	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	537.0	49.4	70.6	53.1	34.2	53.9	65.2

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
RMT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	218.0	42.6	65.8	45.9	—	—	—
RMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	262.0	43.9	67.8	47.4	—	—	—
RMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	262.0	44.9	69.1	48.4	—	—	—
RMT-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	741.0	46.1	69.8	49.8	—	—	—
RMT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	373.0	45.5	69.7	49.3	—	—	—
RMT-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	373.0	46.1	70.4	49.9	—	—	—
RMT-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	852.0	47.2	70.5	51.4	—	—	—
RMT-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	557.0	45.9	70.3	49.8	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
RMT-T	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	33.7	—	—	—	46.4	—	—
RMT-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	937.0	—	—	—	49.8	—	—
RMT-S	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	180.0	—	—	—	49.4	—	—
RMT-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1051.0	—	—	—	52.0	—	—
RMT-B	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	294.0	—	—	—	50.4	—	—
RMT-L	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1241.0	—	—	—	52.8	—	—
RMT-L	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	482.0	—	—	—	51.4	—	—