Heedless Backbones

H-ViT³ Family

Select an option
Results
Parameters (M)
Images / Second
Publication Date
Select an option
---------
Object Detection
Instance Segmentation
Classification
Semantic Segmentation
Panoptic Segmentation
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)
Select an option
----------
Top-1
Top-5
GFLOPs
Select an option
----------
224x224
384x384
512x512
Select an option
Results
Parameters (M)
Images / Second
GFLOPs
Publication Date
Select an option
---------
ImageNet-1k
ImageNet-22k
JFT-300M
JFT-3B
MegData73M
WT++London
WT++
Select an option
----------
Supervised
Sup. + TL
FCMAE
MAE
CL
MAP
Self-Distillation
Select an option
Family
Pretrain Dataset
Classification Resolution
Select an option
----------
Family
Pretrain Method
Classification Resolution
modelparams (m)pretrainfinetuneGFLOPsTop-1
H-ViT³-T29.0IN-1k : Sup. : 300— : — : —4.983.5
H-ViT³-S54.0IN-1k : Sup. : 300— : — : —8.884.4
H-ViT³-B94.0IN-1k : Sup. : 300— : — : —16.784.9
H-ViT³-T29.0IN-1k : Sup. : 300— : — : —4.984.0
H-ViT³-S54.0IN-1k : Sup. : 300— : — : —8.884.9
H-ViT³-B94.0IN-1k : Sup. : 300— : — : —16.785.5
modelparams (m)pretrainfinetunegflopsIN-1k
H-ViT³-T29.0IN-1k : Sup. : 300— : — : —4.983.5/—
H-ViT³-S54.0IN-1k : Sup. : 300— : — : —8.884.4/—
H-ViT³-B94.0IN-1k : Sup. : 300— : — : —16.784.9/—
H-ViT³-T29.0IN-1k : Sup. : 300— : — : —4.984.0/—
H-ViT³-S54.0IN-1k : Sup. : 300— : — : —8.884.9/—
H-ViT³-B94.0IN-1k : Sup. : 300— : — : —16.785.5/—

COCO (val)

modelpretrainheadtraingflopsmAPbAPb50APb75mAPbsmAPbmmAPbl
H-ViT³-TIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12271.047.369.852.3———
H-ViT³-TIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36271.048.971.053.4———
H-ViT³-SIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12349.049.171.553.7———
H-ViT³-SIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36349.050.572.055.5———
H-ViT³-BIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12510.050.071.855.0———
H-ViT³-BIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36510.051.072.155.9———

COCO (val)

modelpretrainheadtraingflopsmAPmAPm50APm75mAPmsmAPmmmAPml
H-ViT³-TIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12271.042.866.846.2———
H-ViT³-TIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36271.044.068.047.5———
H-ViT³-SIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12349.044.168.447.4———
H-ViT³-SIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36349.045.069.148.8———
H-ViT³-BIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12510.044.669.047.7———
H-ViT³-BIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36510.045.369.349.1———

ADE20K (val)

modelpretrainheadtraingflopsmIoUmspAccmsmAccmsmIoUsspAccssmAccss
H-ViT³-TIN-1k : Sup. : 300UPerNetADE20K (train) : 128 : 512946.0———48.0——
H-ViT³-SIN-1k : Sup. : 300UPerNetADE20K (train) : 128 : 5121026.0———50.2——
H-ViT³-BIN-1k : Sup. : 300UPerNetADE20K (train) : 128 : 5121195.0———51.7——