Heedless Backbones

CoCA ViT 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

---------

ImageNet-1k

ImageNet-22k

JFT-300M

JFT-3B

MegData73M

pretrain method

Select an option

----------

Supervised

Sup. + TL

FCMAE

MAE

CL

MAP

Legend

Legend attribute

Select an option

Family

Pretrain Dataset

Pretrain Method

Classification Resolution

Legend attribute (second)

Select an option

----------

Pretrain Dataset

Pretrain Method

Classification Resolution

model	params (m)	pretrain	finetune	GFLOPs	Top-1
CoCA ViT-11M	11.4	IN-1k : Sup. : 300	— : — : —	2.2	82.7
CoCA ViT-21M	20.6	IN-1k : Sup. : 300	— : — : —	4.1	83.6
CoCA ViT-28M	27.8	IN-1k : Sup. : 300	— : — : —	4.9	84.0

model	params (m)	pretrain	finetune	gflops	IN-1k	IN-ReaL	IN-V2	IN-A	IN-R	IN-Sketch
CoCA ViT-11M	11.4	IN-1k : Sup. : 300	— : — : —	2.2	82.7/—	87.8/—	72.9/—	34.1/—	47.8/—	35.5/—
CoCA ViT-21M	20.6	IN-1k : Sup. : 300	— : — : —	4.1	83.6/—	88.3/—	73.7/—	38.9/—	50.1/—	39.6/—
CoCA ViT-28M	27.8	IN-1k : Sup. : 300	— : — : —	4.9	84.0/—	88.4/—	74.0/—	39.8/—	51.1/—	40.2/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
CoCA ViT-21M	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	None	51.8	70.5	56.1	—	—	—
CoCA ViT-28M	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	None	52.2	71.0	56.8	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
CoCA ViT-21M	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	None	44.9	67.8	48.2	—	—	—
CoCA ViT-28M	IN-1k : Sup. : 300	Cascade Mask R-CNN	COCO (train) : 36	None	45.2	68.9	48.8	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
CoCA ViT-21M	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	None	50.8	—	—	—	—	—
CoCA ViT-28M	IN-1k : Sup. : 300	UPerNet	ADE20K (train) : 128 : 512	None	51.3	—	—	—	—	—