deformable attention

Verformbare Aufmerksamkeit

Implementierung von Deformable Attention aus diesem Artikel in Pytorch, was eine Verbesserung gegenüber dem zu sein scheint, was in DETR vorgeschlagen wurde. Die relative Positionseinbettung wurde ebenfalls zur besseren Extrapolation geändert, indem die in SwinV2 vorgeschlagene kontinuierliche Positionseinbettung verwendet wurde.

Installieren

$ pip install deformable-attention

Verwendung

 import torch
from deformable_attention import DeformableAttention

attn = DeformableAttention (
    dim = 512 ,                   # feature dimensions
    dim_head = 64 ,               # dimension per head
    heads = 8 ,                   # attention heads
    dropout = 0. ,                # dropout
    downsample_factor = 4 ,       # downsample factor (r in paper)
    offset_scale = 4 ,            # scale of offset, maximum offset
    offset_groups = None ,        # number of offset groups, should be multiple of heads
    offset_kernel_size = 6 ,      # offset kernel size
)

x = torch . randn ( 1 , 512 , 64 , 64 )
attn ( x ) # (1, 512, 64, 64)

3D verformbare Aufmerksamkeit

 import torch
from deformable_attention import DeformableAttention3D

attn = DeformableAttention3D (
    dim = 512 ,                          # feature dimensions
    dim_head = 64 ,                      # dimension per head
    heads = 8 ,                          # attention heads
    dropout = 0. ,                       # dropout
    downsample_factor = ( 2 , 8 , 8 ),      # downsample factor (r in paper)
    offset_scale = ( 2 , 8 , 8 ),           # scale of offset, maximum offset
    offset_kernel_size = ( 4 , 10 , 10 ),   # offset kernel size
)

x = torch . randn ( 1 , 512 , 10 , 32 , 32 ) # (batch, dimension, frames, height, width)
attn ( x ) # (1, 512, 10, 32, 32)

1d verformbare Aufmerksamkeit für ein gutes Maß

 import torch
from deformable_attention import DeformableAttention1D

attn = DeformableAttention1D (
    dim = 128 ,
    downsample_factor = 4 ,
    offset_scale = 2 ,
    offset_kernel_size = 6
)

x = torch . randn ( 1 , 128 , 512 )
attn ( x ) # (1, 128, 512) 

Zitat

 @misc { xia2022vision ,
    title   = { Vision Transformer with Deformable Attention } , 
    author  = { Zhuofan Xia and Xuran Pan and Shiji Song and Li Erran Li and Gao Huang } ,
    year    = { 2022 } ,
    eprint  = { 2201.00520 } ,
    archivePrefix = { arXiv } ,
    primaryClass = { cs.CV }
}

 @misc { liu2021swin ,
    title   = { Swin Transformer V2: Scaling Up Capacity and Resolution } ,
    author  = { Ze Liu and Han Hu and Yutong Lin and Zhuliang Yao and Zhenda Xie and Yixuan Wei and Jia Ning and Yue Cao and Zheng Zhang and Li Dong and Furu Wei and Baining Guo } ,
    year    = { 2021 } ,
    eprint  = { 2111.09883 } ,
    archivePrefix = { arXiv } ,
    primaryClass = { cs.CV }
}