Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 22, 2017

• fix armv7 assembly

  nihui

  committed
  0564021

  Copy full SHA for 0564021

• unroll outch for convolution 1x1 stride 1

  nihui

  committed
  55ec189

  Copy full SHA for 55ec189

• neon optimize for depthwise convolution 3x3, about 20%~35% speed gain

  nihui

  committed
  57df107

  Copy full SHA for 57df107

Commits on Dec 21, 2017

• fix pooling2x2s2_max_neon stride bug

  Show description for 8222142
  

  water668

  authored and

  nihui

  committed
  8222142

  Copy full SHA for 8222142

Commits on Dec 20, 2017

• mxnet no square convolution

  

  vsooda

  authored and

  nihui

  committed
  e85bebb

  Copy full SHA for e85bebb

Commits on Dec 19, 2017

• strict Werror may cause unexpected compile error

  nihui

  committed
  9e36e2b

  Copy full SHA for 9e36e2b

• warning--

  nihui

  committed
  a240678

  Copy full SHA for a240678

Commits on Dec 18, 2017

• fix mobilenet: add depthwise, fix batch norm (Tencent#218)

  Show description for 02cff84
  

  vsooda

  authored and

  nihui

  committed
  02cff84

  Copy full SHA for 02cff84

• [fix-compile-warnings] fix compiler warnings, and add werror in CMakeLists.txt (Tencent#217)

  Show description for a54f14f
  

  sethbrin

  authored and

  nihui

  committed
  a54f14f

  Copy full SHA for a54f14f

Commits on Dec 16, 2017

• change input param order to w h c, replace caffe MemoryData to Input

  nihui

  committed
  0f52418

  Copy full SHA for 0f52418

• convert non-square convolution pooling param

  nihui

  committed
  abfd3ea

  Copy full SHA for abfd3ea

• non-square kernel stride and padding w h for pooling

  nihui

  committed
  6291396

  Copy full SHA for 6291396

• padding w h in convolution and deconvolution

  nihui

  committed
  bdb70a2

  Copy full SHA for bdb70a2

• non-square convolution and deconvolution kernel stride dilation

  nihui

  committed
  44b4519

  Copy full SHA for 44b4519

Commits on Dec 11, 2017

• fix memcpy error

  

  HustCoderHu

  authored and

  nihui

  committed
  23a3254

  Copy full SHA for 23a3254

Commits on Dec 10, 2017

• add ShuffleChannel layer, only cpp, no arm yet (Tencent#210)

  Show description for 81fb381
  

  HustCoderHu

  authored and

  nihui

  committed
  81fb381

  Copy full SHA for 81fb381

Commits on Dec 8, 2017

• more runtime decisions for winograd path

  nihui

  committed
  964040f

  Copy full SHA for 964040f

Commits on Dec 6, 2017

• fix openmp not found

  

  vsooda

  authored and

  nihui

  committed
  a6d5db8

  Copy full SHA for a6d5db8

Commits on Dec 2, 2017

• enable conv3x3s1 winograd optimization, two paths for small image on armv7 and all for aarch64

  nihui

  committed
  c77ca16

  Copy full SHA for c77ca16

• fix winograd neon2 for aarch64

  nihui

  committed
  f2f7ecd

  Copy full SHA for f2f7ecd

• memcpy for concat

  nihui

  committed
  2630361

  Copy full SHA for 2630361

Commits on Dec 1, 2017

• check clone empty

  nihui

  committed
  a4d2810

  Copy full SHA for a4d2810

• implement external scale blob, support SENet

  nihui

  committed
  25f19c2

  Copy full SHA for 25f19c2

Commits on Nov 30, 2017

• forward reuse forward_inplace routine, reduce binary size with little memcpy overhead in non-light mode

  nihui

  committed
  15ad4df

  Copy full SHA for 15ad4df

Commits on Nov 23, 2017

• use mul for the first multiply, drop accumulator clear instructions, about 5% speed performance gains

  nihui

  committed
  32cd5f2

  Copy full SHA for 32cd5f2

• add squeezenet ssd example

  nihui

  committed
  b800aaa

  Copy full SHA for b800aaa

• fix deconv4x4s2, fix Tencent#202

  nihui

  committed
  d5da0e8

  Copy full SHA for d5da0e8

Commits on Nov 22, 2017

• do not print relu negative slope if it is zero

  nihui

  committed
  96aa5d6

  Copy full SHA for 96aa5d6

• compability improvement for https://github.com/farmingyard/caffe-mobilenet

  nihui

  committed
  8e12bc9

  Copy full SHA for 8e12bc9

• fix unaryop bug (Tencent#200)

  Show description for 429e98c
  

  water668

  authored and

  nihui

  committed
  429e98c

  Copy full SHA for 429e98c

Commits on Nov 21, 2017

• fix top_blob not set (Tencent#199)

  

  huyn

  authored and

  nihui

  committed
  8b9365a

  Copy full SHA for 8b9365a

Commits on Nov 16, 2017

• check dropout has_scale_train

  nihui

  committed
  e8248ec

  Copy full SHA for e8248ec

Commits on Nov 14, 2017

• Update relu_arm.cpp (Tencent#189)

  Show description for f232c1a
  

  Feng137

  authored and

  nihui

  committed
  f232c1a

  Copy full SHA for f232c1a

• Add depthwise Deconvolution. (Tencent#187)

  Show description for 4d59d0a
  

  tedder59

  authored and

  nihui

  committed
  4d59d0a

  Copy full SHA for 4d59d0a

Commits on Nov 12, 2017

• partition dot tiles and reuse kernel register, over 20% improvement for tiny image

  nihui

  committed
  790829b

  Copy full SHA for 790829b

Pagination

Previous

Next