Heedless Backbones

TinyViM 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
TinyViM-S	5.6	IN-1k : Sup. : 300	— : — : —	0.9	79.2
TinyViM-B	11.0	IN-1k : Sup. : 300	— : — : —	1.5	81.2
TinyViM-L	31.7	IN-1k : Sup. : 300	— : — : —	4.72	83.3
TinyViM-S	5.6	IN-1k : Sup. : 1000	— : — : —	0.9	80.3
TinyViM-B	11.0	IN-1k : Sup. : 1000	— : — : —	1.5	82.1
TinyViM-L	31.7	IN-1k : Sup. : 1000	— : — : —	4.7	83.6

model	params (m)	pretrain	finetune	gflops	IN-1k
TinyViM-S	5.6	IN-1k : Sup. : 300	— : — : —	0.9	79.2/—
TinyViM-B	11.0	IN-1k : Sup. : 300	— : — : —	1.5	81.2/—
TinyViM-L	31.7	IN-1k : Sup. : 300	— : — : —	4.72	83.3/—
TinyViM-S	5.6	IN-1k : Sup. : 1000	— : — : —	0.9	80.3/—
TinyViM-B	11.0	IN-1k : Sup. : 1000	— : — : —	1.5	82.1/—
TinyViM-L	31.7	IN-1k : Sup. : 1000	— : — : —	4.7	83.6/—

COCO (val)

model	pretrain	head	train	gflops	mAP^b	AP^b₅₀	AP^b₇₅	mAP^b_s	mAP^b_m	mAP^b_l
TinyViM-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	38.7	61.1	41.3	—	—	—
TinyViM-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	42.3	64.2	46.3	—	—	—
TinyViM-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	44.5	66.4	48.6	—	—	—

COCO (val)

model	pretrain	head	train	gflops	mAP^m	AP^m₅₀	AP^m₇₅	mAP^m_s	mAP^m_m	mAP^m_l
TinyViM-S	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	35.7	57.1	37.9	—	—	—
TinyViM-B	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	38.7	61.1	41.3	—	—	—
TinyViM-L	IN-1k : Sup. : 300	Mask R-CNN	COCO (train) : 12	None	40.7	63.6	43.8	—	—	—

ADE20K (val)

model	pretrain	head	train	gflops	mIoU^ms	pAcc^ms	mAcc^ms	mIoU^ss	pAcc^ss	mAcc^ss
TinyViM-S	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	None	—	—	—	38.9	—	—
TinyViM-B	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	None	—	—	—	41.9	—	—
TinyViM-L	IN-1k : Sup. : 300	Panoptic FPN	ADE20K (train) : 64 : 512	None	—	—	—	44.2	—	—