Heedless Backbones

CMT 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

Classification Resolution

Legend attribute (second)

Select an option

----------

Family

Pretrain Method

Classification Resolution

model	params (m)	pretrain	finetune	GFLOPs	Top-1
CMT-Ti	9.49	IN-1k : Sup. : 800	— : — : —	0.64	79.1
CMT-XS	15.24	IN-1k : Sup. : 300	— : — : —	1.54	81.8
CMT-S	25.14	IN-1k : Sup. : 300	— : — : —	4.04	83.5
CMT-B	45.72	IN-1k : Sup. : 300	— : — : —	9.33	84.5
CMT-L	74.7	IN-1k : Sup. : 300	— : — : —	19.5	84.8

model	params (m)	pretrain	finetune	gflops	IN-1k
CMT-Ti	9.49	IN-1k : Sup. : 800	— : — : —	0.64	79.1/94.5
CMT-XS	15.24	IN-1k : Sup. : 300	— : — : —	1.54	81.8/95.8
CMT-S	25.14	IN-1k : Sup. : 300	— : — : —	4.04	83.5/96.6
CMT-B	45.72	IN-1k : Sup. : 300	— : — : —	9.33	84.5/96.9
CMT-L	74.7	IN-1k : Sup. : 300	— : — : —	19.5	84.8/97.1

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
CMT-S	IN-1k : Sup. : 300	RetinaNet	COCO (train) : 12	231.0	44.3	65.5	47.5	27.1	48.3	59.1
CMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	249.0	44.6	66.8	48.9	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
CMT-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	249.0	40.7	63.9	43.4	—	—	—