Heedless Backbones

BiFormer 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

CL

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
BiFormer-T	13.1	IN-1k : Sup. : 300	— : — : —	2.2	81.4
BiFormer-S	26.0	IN-1k : Sup. : 300	— : — : —	4.5	83.8
BiFormer-S	26.0	IN-1k : Sup. + TL : 300	— : — : —	4.5	84.3
BiFormer-B	57.0	IN-1k : Sup. : 300	— : — : —	9.8	84.3
BiFormer-B	57.0	IN-1k : Sup. + TL : 300	— : — : —	9.8	85.4

model	params (m)	pretrain	finetune	gflops	IN-1k
BiFormer-T	13.1	IN-1k : Sup. : 300	— : — : —	2.2	81.4/—
BiFormer-S	26.0	IN-1k : Sup. : 300	— : — : —	4.5	83.8/—
BiFormer-S	26.0	IN-1k : Sup. + TL : 300	— : — : —	4.5	84.3/—
BiFormer-B	57.0	IN-1k : Sup. : 300	— : — : —	9.8	84.3/—
BiFormer-B	57.0	IN-1k : Sup. + TL : 300	— : — : —	9.8	85.4/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
BiFormer-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	45.9	66.9	49.4	30.2	49.6	61.7
BiFormer-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	47.8	69.8	52.3	—	—	—
BiFormer-B	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	None	47.1	68.5	50.4	31.3	50.8	62.6
BiFormer-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	48.6	70.5	53.8	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
BiFormer-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	43.2	66.8	46.5	—	—	—
BiFormer-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	43.7	67.6	47.1	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
BiFormer-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	None	50.8	—	—	49.8	—	—
BiFormer-S	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 128 : 512	None	—	—	—	48.9	—	—
BiFormer-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	None	51.7	—	—	51.0	—	—
BiFormer-B	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 128 : 512	None	—	—	—	49.9	—	—