ConvNeXt 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)

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

RetinaNet

DINO

HTC++

HTC

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

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
ConvNeXt-L	198.0	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1354.0	54.8
ConvNeXt-B	89.0	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	964.0	52.7
ConvNeXt-B	89.0	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	964.0	54.0
ConvNeXt-S	50.0	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	827.0	51.9
ConvNeXt-T	29.0	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	741.0	50.4
ConvNeXt-T	29.0	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	262.0	46.2
ConvNeXt-XL	350.0	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1898.0	55.2

model	params (m)	pretrain	finetune	gflops	IN-1k	IN-V2	IN-Sketch	IN-C↓	IN-A	IN-C-bar↓	IN-R
ConvNeXt-L	198.0	IN-1k : Sup. : 300	— : — : —	34.4	84.3/—	74.0/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-L	198.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	101.0	85.5/—	75.3/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-L	198.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	34.4	86.6/—	76.6/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-L	198.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	101.0	87.5/—	77.7/—	52.8/—	40.2/—	65.5/—	29.9/—	66.7/—
ConvNeXt-B	89.0	IN-1k : Sup. : 300	— : — : —	15.4	83.8/—	73.4/—	38.2/—	46.8/—	36.7/—	34.4/—	51.3/—
ConvNeXt-B	89.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	45.0	85.1/—	74.7/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-B	89.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	15.4	85.8/—	75.6/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-B	89.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	45.1	86.8/—	76.6/—	51.6/—	43.1/—	62.3/—	30.7/—	64.9/—
ConvNeXt-S	50.0	IN-1k : Sup. : 300	— : — : —	8.7	83.1/—	—/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-S	50.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	8.7	84.6/—	—/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-S	50.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	25.5	85.8/—	—/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-T	29.0	IN-1k : Sup. : 300	— : — : —	4.5	82.1/—	—/—	33.8/—	53.2/—	24.2/—	40.0/—	47.2/—
ConvNeXt-T	29.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	4.5	82.9/—	—/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-T	29.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	13.1	84.1/—	—/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-XL	350.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	60.9	87.0/—	77.0/—	—/—	—/—	—/—	—/—	—/—
ConvNeXt-XL	350.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	179.0	87.8/—	77.7/—	55.0/—	38.8/—	69.3/—	27.1/—	68.2/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
ConvNeXt-L	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1354.0	54.8	73.8	59.8	—	—	—
ConvNeXt-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	964.0	52.7	71.3	57.2	—	—	—
ConvNeXt-B	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	964.0	54.0	73.1	58.8	—	—	—
ConvNeXt-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	827.0	51.9	70.8	56.5	—	—	—
ConvNeXt-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	741.0	50.4	69.1	54.8	—	—	—
ConvNeXt-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	262.0	46.2	67.9	50.8	—	—	—
ConvNeXt-XL	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1898.0	55.2	74.2	59.9	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
ConvNeXt-L	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1354.0	47.6	71.3	51.7	—	—	—
ConvNeXt-B	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	964.0	45.6	68.9	49.5	—	—	—
ConvNeXt-B	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	964.0	46.9	70.6	51.3	—	—	—
ConvNeXt-S	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	827.0	45.0	68.4	49.1	—	—	—
ConvNeXt-T	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	741.0	43.7	66.5	47.3	—	—	—
ConvNeXt-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 36	262.0	41.7	65.0	44.9	—	—	—
ConvNeXt-XL	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1898.0	47.7	71.6	52.2	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
ConvNeXt-L	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 128 : 640	2458.0	53.7	—	—	53.2	—	—
ConvNeXt-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1170.0	49.9	—	—	49.1	—	—
ConvNeXt-B	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 128 : 640	1828.0	53.1	—	—	52.6	—	—
ConvNeXt-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1027.0	49.6	—	—	48.7	—	—
ConvNeXt-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	939.0	46.7	—	—	46.0	—	—
ConvNeXt-XL	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 128 : 640	3335.0	54.0	—	—	53.6	—	—