HorNet 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

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
HorNet-T-7x7	22.0	IN-1k : Sup. : 300	— : — : —	4.0	82.8
HorNet-T-GF	23.0	IN-1k : Sup. : 300	— : — : —	3.9	83.0
HorNet-S-7x7	50.0	IN-1k : Sup. : 300	— : — : —	8.8	83.8
HorNet-S-GF	50.0	IN-1k : Sup. : 300	— : — : —	8.7	84.0
HorNet-B-7x7	87.0	IN-1k : Sup. : 300	— : — : —	15.6	84.2
HorNet-B-7x7	87.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	45.8	85.3
HorNet-B-GF	88.0	IN-1k : Sup. : 300	— : — : —	15.5	84.3
HorNet-B-GF	88.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	45.4	85.6
HorNet-L-7x7	195.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	34.8	86.8
HorNet-L-7x7	195.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	102.3	87.6
HorNet-L-GF	196.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	34.6	87.0
HorNet-L-GF	196.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	101.8	87.7

model	params (m)	pretrain	finetune	gflops	IN-1k
HorNet-T-7x7	22.0	IN-1k : Sup. : 300	— : — : —	4.0	82.8/—
HorNet-T-GF	23.0	IN-1k : Sup. : 300	— : — : —	3.9	83.0/—
HorNet-S-7x7	50.0	IN-1k : Sup. : 300	— : — : —	8.8	83.8/—
HorNet-S-GF	50.0	IN-1k : Sup. : 300	— : — : —	8.7	84.0/—
HorNet-B-7x7	87.0	IN-1k : Sup. : 300	— : — : —	15.6	84.2/—
HorNet-B-7x7	87.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	45.8	85.3/—
HorNet-B-GF	88.0	IN-1k : Sup. : 300	— : — : —	15.5	84.3/—
HorNet-B-GF	88.0	IN-1k : Sup. : 300	IN-1k : 30 : 384	45.4	85.6/—
HorNet-L-7x7	195.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	34.8	86.8/—
HorNet-L-7x7	195.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	102.3	87.6/—
HorNet-L-GF	196.0	IN-22k : Sup. : 90	IN-1k : 30 : 224	34.6	87.0/—
HorNet-L-GF	196.0	IN-22k : Sup. : 90	IN-1k : 30 : 384	101.8	87.7/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
HorNet-T-7x7	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	730.0	51.7	—	—	—	—	—
HorNet-T-GF	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	728.0	52.4	—	—	—	—	—
HorNet-S-7x7	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	830.0	52.7	—	—	—	—	—
HorNet-S-7x7	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	351.0	47.1	—	—	—	—	—
HorNet-S-GF	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	827.0	53.3	—	—	—	—	—
HorNet-B-7x7	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	969.0	53.3	—	—	—	—	—
HorNet-B-GF	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	965.0	54.0	—	—	—	—	—
HorNet-L-7x7	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1363.0	55.4	—	—	—	—	—
HorNet-L-GF	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1358.0	56.0	—	—	—	—	—
HorNet-L-GF	IN-22k : Sup. : 90	HTC++	COCO (train) : 36	None	58.1	—	—	—	—	—
HorNet-L-GF	IN-22k : Sup. : 90	DINO	COCO (train) : 36	None	59.2	—	—	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
HorNet-T-7x7	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	730.0	44.8	—	—	—	—	—
HorNet-T-GF	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	728.0	45.6	—	—	—	—	—
HorNet-S-7x7	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	830.0	45.6	—	—	—	—	—
HorNet-S-7x7	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	351.0	42.2	—	—	—	—	—
HorNet-S-GF	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	827.0	46.3	—	—	—	—	—
HorNet-B-7x7	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	969.0	46.1	—	—	—	—	—
HorNet-B-GF	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	965.0	46.9	—	—	—	—	—
HorNet-L-7x7	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1363.0	48.0	—	—	—	—	—
HorNet-L-GF	IN-22k : Sup. : 90	Cascade Mask R-CNN	COCO (train) : 36	1358.0	48.6	—	—	—	—	—
HorNet-L-GF	IN-22k : Sup. : 90	HTC++	COCO (train) : 36	None	50.5	—	—	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
HorNet-T-7x7	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	926.0	48.9	—	—	48.1	—	—
HorNet-T-GF	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	924.0	49.3	—	—	49.2	—	—
HorNet-S-7x7	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1030.0	49.8	—	—	49.2	—	—
HorNet-S-GF	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1027.0	50.5	—	—	50.0	—	—
HorNet-B-7x7	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1174.0	50.5	—	—	50.0	—	—
HorNet-B-GF	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1171.0	50.9	—	—	50.5	—	—
HorNet-L-7x7	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 128 : 640	2473.0	54.5	—	—	54.1	—	—
HorNet-L-GF	IN-22k : Sup. : 90	UPerNet	ADE20K (train) : 128 : 640	2465.0	55.2	—	—	55.0	—	—
HorNet-L-GF	IN-22k : Sup. : 90	Mask2Former	ADE20K (train) : 128 : 640	None	57.9	—	—	57.5	—	—