Heedless Backbones

FAT 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
FAT-B0	4.5	IN-1k : Sup. : 300	— : — : —	0.72	77.6
FAT-B1	7.8	IN-1k : Sup. : 300	— : — : —	1.2	80.1
FAT-B2	13.5	IN-1k : Sup. : 300	— : — : —	2.0	81.9
FAT-B3	29.0	IN-1k : Sup. : 300	— : — : —	4.4	83.6
FAT-B3-ST	29.0	IN-1k : Sup. : 300	— : — : —	4.7	83.0

model	params (m)	pretrain	finetune	gflops	IN-1k
FAT-B0	4.5	IN-1k : Sup. : 300	— : — : —	0.72	77.6/—
FAT-B1	7.8	IN-1k : Sup. : 300	— : — : —	1.2	80.1/—
FAT-B2	13.5	IN-1k : Sup. : 300	— : — : —	2.0	81.9/—
FAT-B3	29.0	IN-1k : Sup. : 300	— : — : —	4.4	83.6/—
FAT-B3-ST	29.0	IN-1k : Sup. : 300	— : — : —	4.7	83.0/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
FAT-B0	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	40.4	61.6	42.7	24.0	44.3	53.1
FAT-B0	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	40.8	63.3	44.2	—	—	—
FAT-B1	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	42.5	64.0	45.1	26.9	46.0	56.7
FAT-B1	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	43.3	65.6	47.4	—	—	—
FAT-B2	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	44.0	65.2	47.2	27.5	47.7	58.8
FAT-B2	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	45.2	67.9	49.0	—	—	—
FAT-B3	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	45.9	66.9	49.5	29.3	50.1	60.9
FAT-B3	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	47.6	69.7	52.3	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
FAT-B0	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	37.7	60.2	40.0	—	—	—
FAT-B1	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	39.6	61.9	42.8	—	—	—
FAT-B2	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	41.3	64.6	44.0	—	—	—
FAT-B3	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	43.1	66.4	46.2	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
FAT-B0	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	25.0	—	—	—	41.5	—	—
FAT-B1	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	27.5	—	—	—	42.9	—	—
FAT-B2	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	32.2	—	—	—	45.4	—	—
FAT-B3	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	179.0	—	—	—	48.9	—	—
FAT-B3	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	936.0	50.7	—	—	49.6	—	—