685 Commits (a3a2548aa28a9ff7924f76d5085fabe94de79ddb)
 

Author SHA1 Message Date
  Zexin, Hu 81fb3818a5 add ShuffleChannel layer, only cpp, no arm yet (#210) 8 years ago
  nihuini 964040fe3c more runtime decisions for winograd path 8 years ago
  vsooda a6d5db83b0 fix openmp not found 8 years ago
  nihui c77ca16468 enable conv3x3s1 winograd optimization, two paths for small image on armv7 and all for aarch64 8 years ago
  nihui f2f7ecd2ec fix winograd neon2 for aarch64 8 years ago
  nihui 26303615a6 memcpy for concat 8 years ago
  nihuini a4d28107f4 check clone empty 8 years ago
  nihuini 25f19c2009 implement external scale blob, support SENet 8 years ago
  nihui 15ad4dfb9f forward reuse forward_inplace routine, reduce binary size with little memcpy overhead in non-light mode 8 years ago
  nihui 32cd5f2a5c use mul for the first multiply, drop accumulator clear instructions, about 5% speed performance gains 8 years ago
  nihuini b800aaa30b add squeezenet ssd example 8 years ago
  nihuini d5da0e84ba fix deconv4x4s2, fix #202 8 years ago
  nihuini 96aa5d61af do not print relu negative slope if it is zero 8 years ago
  nihuini 8e12bc916d compability improvement for https://github.com/farmingyard/caffe-mobilenet 8 years ago
  wind19870521 429e98c91c fix unaryop bug (#200) 8 years ago
  huyn 8b9365a68c fix top_blob not set (#199) 8 years ago
  nihuini e8248ecab0 check dropout has_scale_train 8 years ago
  azrael0fog f232c1a6c5 Update relu_arm.cpp (#189) 8 years ago
  tedder59 4d59d0afda Add depthwise Deconvolution. (#187) 8 years ago
  nihui 790829bc62 partition dot tiles and reuse kernel register, over 20% improvement for tiny image 8 years ago
  nihuini a3be17eb7e special path for 1x1xc innerproduct 8 years ago
  nihuini 50d591cb50 softmax inplace 8 years ago
  peng 39445b5233 no memcpy for small size copy_cut_border/copy_make_boder 8 years ago
  彭 a86cc8f620 memcpy optimize copy_cut_border/copy_make_boder (#179) 8 years ago
  nihuini 1635460c50 fix batchnorm weights conversion 8 years ago
  nihuini 20f49b677c move caffe2ncnn to subdir, build mxnet2ncnn 8 years ago
  nihuini 2d2d1666e2 no input shape 8 years ago
  nihuini d99f9d9ac3 implement softmax on vector and image 8 years ago
  nihuini b58a63f06a special case for SoftmaxOutput 8 years ago
  nihuini 059edd5488 improve compability with old model, wip... 8 years ago
  nihuini bce154ce4e add pytorch tool readme 8 years ago
  nihuini ba734c412a recognize Input and Split, write ncnn param 8 years ago
  nihuini ef624a0850 parse mxnet json param 8 years ago
  liuchang 794a31987b Add autotest folder. 8 years ago
  liuchang ac3b4768aa fix the missing header file for visual studio. 8 years ago
  nihuini ff3c03cfb1 q9 is useless 8 years ago
  nihuini 8cfd02d633 Merge branch 'master' of https://github.com/Tencent/ncnn 8 years ago
  nihuini 9a55404c72 fix dot on aarch64, still needs improvement ... 8 years ago
  nihui eea3ca577a disable winograd atm ... 8 years ago
  nihui 0385d8e8ad implement winograd64 optimization for convolution 3x3s1 8 years ago
  nihui e84891f6d4 add how to build 8 years ago
  nihui 20b1330cdb fix lrn within channel 8 years ago
  nihuini fd6af0f2fe fix opencv detection, second try 8 years ago
  nihui d8193dddcc do not fail on opencv 2.x 8 years ago
  nihui 70ad4c1275 fix param files for dropout fix 8 years ago
  nihui cee9f54016 fix dropout 8 years ago
  nihui 8e490d4b68 fix array parsing, first try 8 years ago
  nihui 908a8f48d2 assign same size 8 years ago
  nihui 4fabb4c274 more encapsulation for paramdict 8 years ago
  nihui d6c833a715 scale_data is vec 8 years ago