DAT 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

---------

Cityscapes (val)

Cityscapes (test)

ADE20K (val)

ADE20K (test)

PASCAL VOC 2007 (val)

PASCAL VOC 2007 (test)

metric

Select an option

mIoU^ms

pAcc^ms

mAcc^ms

mIoU^ss

pAcc^ss

mAcc^ss

GFLOPs

head

Select an option

---------

UPerNet

Mask2Former

Panoptic FPN

SETR

resolution

Select an option

----------

512x2048

640x2560

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

MAP

Legend

Legend attribute

Select an option

Family

Pretrain Dataset

Semantic Segmentation Head

Semantic Segmentation Resolution

Semantic Segmentation Training Epochs

Legend attribute (second)

Select an option

----------

Family

Pretrain Method

Semantic Segmentation Head

Semantic Segmentation Resolution

Semantic Segmentation Training Epochs

model	params (m)	pretrain	head	train	GFLOPs	mIoU^ms
DAT-T	29.0	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	957.0	46.44
DAT-T	29.0	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 32 : 512	198.0	44.22
DAT-S	50.0	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1079.0	49.84
DAT-S	50.0	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 32 : 512	320.0	48.46
DAT-B	88.0	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1212.0	50.55
DAT-B	88.0	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 32 : 512	481.0	49.01

model	params (m)	pretrain	finetune	gflops	IN-1k
DAT-T	29.0	IN-1k : Sup. : 300	— : — : —	4.6	82.0/—
DAT-S	50.0	IN-1k : Sup. : 300	— : — : —	9.0	83.7/—
DAT-B	88.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	49.8	84.8/—
DAT-B	88.0	IN-1k : Sup. : 300	— : — : —	15.8	84.0/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
DAT-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	253.0	42.8	64.4	45.2	28.0	45.8	57.8
DAT-T	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 36	253.0	45.6	67.2	48.5	31.3	49.1	60.8
DAT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	272.0	44.4	67.6	48.5	28.3	47.5	58.5
DAT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	272.0	47.1	69.2	51.6	32.0	50.3	61.0
DAT-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 12	750.0	49.1	68.2	52.9	31.2	52.4	65.1
DAT-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	750.0	51.3	70.1	55.8	34.1	54.6	66.9
DAT-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	359.0	45.7	67.7	48.5	30.5	49.3	61.3
DAT-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 36	359.0	47.9	69.6	51.2	32.3	51.8	63.4
DAT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	378.0	47.1	69.9	51.5	30.5	50.1	62.1
DAT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	378.0	49.0	70.9	53.8	32.7	52.6	64.0
DAT-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	857.0	52.7	71.7	57.2	37.3	56.3	68.0
DAT-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	1003.0	53.0	71.9	57.6	36.0	56.8	69.1

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
DAT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	272.0	40.4	64.2	43.1	23.9	43.8	55.5
DAT-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	272.0	42.4	66.1	45.5	27.2	45.8	57.1
DAT-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 12	750.0	42.5	65.4	45.8	25.2	45.9	58.6
DAT-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	750.0	44.5	67.5	48.1	27.9	47.9	60.3
DAT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	378.0	42.5	66.7	45.4	25.5	45.8	58.5
DAT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	378.0	44.0	68.0	47.5	27.8	47.7	59.5
DAT-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	857.0	45.5	69.1	49.3	30.2	49.2	60.9
DAT-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	1003.0	45.8	69.3	49.5	29.2	49.5	61.9

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
DAT-T	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 32 : 512	198.0	44.22	—	—	42.56	—	54.72
DAT-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	957.0	46.44	—	—	45.54	—	57.95
DAT-S	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 32 : 512	320.0	48.46	—	—	46.08	—	58.17
DAT-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1079.0	49.84	—	—	48.31	—	60.44
DAT-B	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 32 : 512	481.0	49.01	—	—	47.02	—	59.47
DAT-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1212.0	50.55	—	—	49.38	—	61.82