Muti-scale related papers

Omni-Scale

说明问题

homogeneous scales，多个尺度的feature分别比较。
heterogeneous scales, 多个尺度的feature组合后有辨识度。（白T+logo）

创新点

Omni-Scale

多个感受野的分支相加，保证多尺度

AG gate

不同分支共享一个SEBlock

Residual 解释

The shortcut connection
allows features at smaller scales learned in the current layer
to be preserved effectively in the next layers, thus enabling
the final features to capture a whole range of spatial scales.

Light weight

Depthwise Separable Convolutions:depthwise kernel&pointwise kernel

对比对象

• 网络结构： Multi-stream:Inception ResNeXt /SENet
• re-id paper :
  Person Re-Identification by Deep Learning Multi-Scale Representations 多分支带来多尺度
  Multi-scaleDeepLearningArchitecturesforPersonRe-identification

Ablation study

1.ResNeXt-like
2.aggression method:cocat/add/separate ag/stream-wise AG/learned-and-fixed gates
3.scale:1/2/3

可视化

• 可视化AG GATE
  As the gating
  vectors produced by the AG inherently encode the way
  how the omni-scale feature streams are aggregated, we can
  understand what the AG sub-network has learned by visualising
  images of similar gating vectors. To this end, we
  concatenate the gating vectors of four streams in the last
  bottleneck, perform k-means clustering on test images of
  Mix4, and select top-15 images closest to the cluster centres.
  Fig. 5 shows four example clusters where images
  within the same cluster exhibit similar patterns, i.e., combinations
  of global-scale and local-scale appearance.
• 可视化learned features
  visualise
  the activations of the last convolutional feature maps
  to investigate where the network focuses on to extract features

Multi-Level Factorisation Net for Person Re-Identification

问题描述

low-stage ：lowlevel semantic attributes 衣服颜色
high-stage: highlevel semantic attributes 携带东西、性别

创新点

ReXnet+组内选择FACTOR 选择向量为fsm 表示某一层的选取情况
fsm与output concact

对比对象

resNeXt Mixture-of-expert

ablation

ResNet ResNeXt MLFN-Fusion:without fusion of FS and output

可视化

we rank each element
of the FSM output vector, denoted as Sn;i, with all
testing samples in Market-1501 [55] as inputs. Person images
with the highest and lowest twenty values of each Sn;i
are recorded.