Heedless Backbones

Pale 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
Pale-T	22.0	IN-1k : Sup. : 300	— : — : —	4.2	83.4
Pale-S	48.0	IN-1k : Sup. : 300	— : — : —	9.0	84.3
Pale-B	85.0	IN-1k : Sup. : 300	— : — : —	15.6	84.9
Pale-T	22.0	IN-1k : Sup. + TL : 300	— : — : —	4.2	84.2
Pale-S	48.0	IN-1k : Sup. + TL : 300	— : — : —	9.0	85.0
Pale-B	85.0	IN-1k : Sup. + TL : 300	— : — : —	15.6	85.8

model	params (m)	pretrain	finetune	gflops	IN-1k
Pale-T	22.0	IN-1k : Sup. : 300	— : — : —	4.2	83.4/—
Pale-S	48.0	IN-1k : Sup. : 300	— : — : —	9.0	84.3/—
Pale-B	85.0	IN-1k : Sup. : 300	— : — : —	15.6	84.9/—
Pale-T	22.0	IN-1k : Sup. + TL : 300	— : — : —	4.2	84.2/—
Pale-S	48.0	IN-1k : Sup. + TL : 300	— : — : —	9.0	85.0/—
Pale-B	85.0	IN-1k : Sup. + TL : 300	— : — : —	15.6	85.8/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
Pale-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	306.0	47.4	69.2	52.3	—	—	—
Pale-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	432.0	48.4	70.4	53.2	—	—	—
Pale-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	595.0	49.3	71.2	54.1	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
Pale-T	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	306.0	42.7	66.3	46.2	—	—	—
Pale-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	432.0	43.7	67.7	47.1	—	—	—
Pale-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	595.0	44.2	68.1	47.8	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
Pale-T	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	996.0	51.2	—	—	50.4	—	—
Pale-S	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1135.0	52.2	—	—	51.2	—	—
Pale-B	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	1311.0	53.0	—	—	52.2	—	—