Heedless Backbones

TransNeXt 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

---------

COCO (val)

COCO (test)

Cityscapes (val)

Cityscapes (test)

ADE20K (val)

ADE20K (test)

metric

Select an option

----------

mAP

AP₅₀

AP₇₅

mAP_s

mAP_m

mAP_l

GFLOPs

head

Select an option

---------

Mask R-CNN

Cascade Mask R-CNN

Mask2Former

HTC++

HTC

Panoptic FPN

axis

Select an option

Results

Parameters (M)

Images / Second

GFLOPs

Publication Date

Filters

pretrain dataset

Select an option

---------

JFT-3B

ImageNet-1k

ImageNet-22k

JFT-300M

pretrain method

Select an option

----------

Supervised

FCMAE

MAE

CL

Legend

Legend attribute

Select an option

Family

Pretrain Dataset

Pretrain Method

Instance Head

Instance Training Epochs

Legend attribute (second)

Select an option

----------

Family

Pretrain Method

Instance Head

Instance Training Epochs

model	params (m)	pretrain	head	train	GFLOPs	mAP
TransNeXt-T	28.2	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	356.0	44.6
TransNeXt-S	49.7	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	516.0	45.5
TransNeXt-B	89.7	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	728.0	45.9

model	params (m)	pretrain	finetune	gflops	IN-1k	IN-C↓	IN-A	IN-R	IN-Sketch	IN-V2
TransNeXt-Micro	12.8	IN-1k : Sup. : 300	— : — : —	2.7	82.5/—	50.8/—	29.9/—	45.8/—	33.0/—	72.6/—
TransNeXt-T	28.2	IN-1k : Sup. : 300	— : — : —	5.7	84.0/—	46.5/—	39.9/—	49.6/—	37.6/—	73.8/—
TransNeXt-S	49.7	IN-1k : Sup. : 300	— : — : —	10.3	84.7/—	43.9/—	47.1/—	52.5/—	39.7/—	74.8/—
TransNeXt-S	49.7	IN-1k : Sup. : 300	IN-1k : 5 : 384	32.1	86.0/—	—/—	58.3/—	56.4/—	43.2/—	76.8/—
TransNeXt-B	89.7	IN-1k : Sup. : 300	— : — : —	18.4	84.8/—	43.5/—	50.6/—	53.9/—	41.4/—	75.1/—
TransNeXt-B	89.7	IN-1k : Sup. : 300	IN-1k : 5 : 384	56.3	86.2/—	—/—	61.6/—	57.7/—	44.7/—	77.0/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
TransNeXt-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	356.0	49.9	71.5	54.9	—	—	—
TransNeXt-T	IN-1k : Sup. : 300	DINO	COCO (train) : 12	None	55.7	—	—	—	—	—
TransNeXt-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	516.0	51.1	72.6	56.2	—	—	—
TransNeXt-S	IN-1k : Sup. : 300	DINO	COCO (train) : 12	None	56.6	—	—	—	—	—
TransNeXt-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	728.0	51.7	73.2	56.9	—	—	—
TransNeXt-B	IN-1k : Sup. : 300	DINO	COCO (train) : 12	None	57.1	—	—	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
TransNeXt-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	356.0	44.6	68.6	48.1	—	—	—
TransNeXt-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	516.0	45.5	69.8	49.1	—	—	—
TransNeXt-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	728.0	45.9	70.5	49.7	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
TransNeXt-T	IN-1k : Sup. : 300	Mask2Former	ADE20K (train) : 128 : 512	None	—	—	—	53.4	—	—
TransNeXt-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	978.0	51.7	—	—	51.1	—	—
TransNeXt-S	IN-1k : Sup. : 300	Mask2Former	ADE20K (train) : 128 : 512	None	—	—	—	54.1	—	—
TransNeXt-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1089.0	52.8	—	—	52.5	—	—
TransNeXt-B	IN-1k : Sup. : 300	Mask2Former	ADE20K (train) : 128 : 512	None	—	—	—	54.7	—	—
TransNeXt-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1268.0	53.7	—	—	53.0	—	—