Proto commits in happynear/caffe-windows

These commits are when the Protocol Buffers files have changed: (only the last 100 relevant commits are shown)

2018-05-28

anneal add

2018-02-06

margin determined by radial removal.

2018-02-01

label specific add annealing

The documentation is generated from this commit.

2018-01-17

label specific add

2018-01-16

layers for AM-Softmax

softmax margin

2018-01-11

auto-tune

2017-12-20

pass bp

2017-12-11

scale for angle

2017-12-10

add some experimental layers

2017-09-14

feature decay loss

label smoothing

2017-09-11

weighted mean

2017-09-10

power

2017-09-08

label specific affine

2017-09-06

split forward and backward in channel scale layer.

2017-09-05

random erasing

soft margin

2017-09-04

auto_tune

label specific margin

2017-09-02

normalize layer fix gradient(don't use this parameter).

2017-08-31

x / norm(x) * norm(x)

2017-08-28

focal loss for softmax

label specific power

2017-08-26

focal loss from https://github.com/sciencefans/Focal-Loss

2017-08-15

label smoothing

2017-08-12

change rescale strategy

2017-08-11

feature incay add force incay.

2017-08-10

base should be 0

add scaling lambda to label specified rescale layer

label specific rescale compatible for innerproduct.

2017-07-21

SphereFace

2017-07-06

Revert "merge ms to hog"

2017-06-15

resize layer

2017-06-07

Merge branch 'master' of https://github.com/bvlc/caffe into ms # Conflicts: # python/caffe/_caffe.cpp # src/caffe/layers/batch_norm_layer.cpp # src/caffe/layers/batch_norm_layer.cu # src/caffe/layers/cudnn_conv_layer.cpp # src/caffe/layers/cudnn_relu_layer.cpp # src/caffe/proto/caffe.proto

2017-06-04

ordinal regression

2017-05-03

bn layer update

2017-04-27

bn layer

2017-04-21

mae for contrastive loss

2017-04-14

Merge pull request #3855 from shaibagon/upgrade_infogain InfogainLoss layer can normalize, ignore, and more

2017-03-27

Remove missed legacy parallel code

2017-03-22

Clarify batch norm parameter documentation.

second order operation for eltwise layer

infimum loss layer, label specific rescale layer

2017-03-18

large margin softmax

2017-03-17

Merge branch 'master' of https://github.com/bvlc/caffe into ms # Conflicts: # matlab/+caffe/private/caffe_.cpp # python/caffe/_caffe.cpp

2017-03-06

make dropout's scale consistent with normalize layer.

2017-03-03

get cudnn_batch_norm layer from nvidia-caffe

2017-02-25

default positive margin set to 0

2017-02-24

add exp weight to contrastive_loss_layer

add min_negative param to nca

2017-02-19

truncation layer for regression

2017-02-16

Merge pull request #5210 from ftokarev/patches Obsolete reference to `bool solver` in caffe.proto

2017-02-14

crop_h crop_w

2017-02-12

clip weight

2017-02-09

permute_layer from SSD

add cut_label parameter to multi_label_image_data for temporal training.

2017-02-07

multi label image data layer add balance

deconv_layer add parameter shape_offset

2017-02-06

add clamp weights to solver

2017-02-05

batch_contrastive_loss_layer

2017-02-04

add turn point to smooth L1 layer

2017-01-30

batch_norm add parameters: disable_mean, disable_variance.

2017-01-29

add positive_first parameter to general contrastive layer.

2017-01-27

if positive distance beyond a upper bound, optimize positive distance only.

pairwise layer.

2017-01-24

general_triplet_layer

2017-01-21

Update a comment in caffe.proto

2017-01-19

add filler to parameter layer.

proposal and psroi layer from RFCN

2017-01-18

Merge branch 'master' of https://github.com/bvlc/caffe into ms # Conflicts: # matlab/+caffe/private/caffe_.cpp # python/caffe/_caffe.cpp # src/caffe/layers/base_data_layer.cpp # src/caffe/layers/base_data_layer.cu

hard positive rename.

2017-01-17

only optimize the hardest sample.

2017-01-15

ignore outlier.

2017-01-06

Switched multi-GPU to NCCL

balance classes by over-sampling.

2017-01-05

general triplet loss(not finished)

2016-12-30

general contrastive loss add positive/negative weights

2016-12-29

add normalization parameter to innerproduct layer.

2016-12-09

add min_is_better param to accuracy layer.

2016-12-07

refact normalize

2016-12-05

write gpu_kernels for normalize layer.

2016-12-04

Merge branch 'ms' of https://github.com/happynear/caffe-windows into ms

initial inner_distance_layer

2016-11-27

fix many typos by using codespell

2016-11-17

sigmoid cross-entropy loss: normalize loss by different schemes sig-ce loss handles all the same normalizations as the softmax loss; refer to #3296 for more detail. this preserves the default normalization for sig-ce loss: batch size.

2016-11-07

add L1 distance to center loss

2016-11-06

flip layer

2016-10-29

Merge remote-tracking branch 'bvlc/master' into ms

2016-10-28

fix transpose param in memory data layer, trying multi-thread predicting.

2016-10-15

center loss

2016-09-27

cascade cnn

2016-08-03

some fix on fcn data & accuracy layer

2016-07-27

fcn data change, fcn accuracy, softmax cutting point.

Add RecurrentLayer: an abstract superclass for other recurrent layer types

2016-06-28

add default value for rms_decay

2016-06-18

Merge branch 'master' of https://github.com/BVLC/caffe into ms Conflicts: scripts/travis/travis_install.sh src/caffe/layers/memory_data_layer.cpp

2016-06-01

Add RecurrentLayer: an abstract superclass for other recurrent layer types

2016-05-21

matcaffe output folder, def pool, io size fix for VGG model.

2016-05-10

Merge bvlc/windows@{2016-05-10} into master