Proto commits in bharath272/sds_eccv2014

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

2014-07-07

renaming

The documentation is generated from this commit.

2014-06-23

all finetuning

2014-06-09

move stuff into caffe directory to start anew

2014-05-09

make solver_mode an enum with CPU and GPU -- fully backwards compatible with old 0/1 style

2014-05-02

add MemoryDataLayer for reading input from contiguous blocks of memory

2014-04-30

note the last added layer/params in caffe.proto to prevent conflicts The current scheme does not actually prevent conflicts, since three-way merge will accept simultaneous changes that agree on the next number. This commit fixes this by explicitly noting the last layer added.

2014-04-25

add HingeLossLayer for one-vs-all hinge loss This layer implements a "one-vs-all" hinge loss, (1/n) sum_ij max(0, 1 - y_ij x_ij), with bottom blob x_ij (i ranging over examples and j over classes), and y_ij = +1/-1 indicating the label. No regularization is included, since regularization is done via weight decay or using the parameters of another layer. The gradient is taken to be zero at the hinge point. This commit only provides the CPU implementation.

2014-04-23

add random_seed field to SolverParameter and have solver use it -- already works for lenet, doesn't work for imagenet w/ rand() calls

2014-04-16

change to correct next layer id for merge

2014-04-15

make solver able to compute and display test loss

add mnist autoencoder example necessities (sigmoid cross entropy loss layer, sparse gaussian filler)

2014-04-08

update proto field IDs from placeholder values

add padding for average pooling

add LRN within map layer and dependencies (eltwise product and power)

merge LRNMapLayer into LRNLayer with norm_region proto field

minor polishing

use average pool instead of conv

2014-03-28

some post rebase fixes -- copyright, hdf5_output layer (still need to incorporate into util/upgrade_proto)

cleaner version of refactoring with fields added to LayerConnection (which retains an optional V0LayerParameter field for legacy support) and LayerConnection renamed to LayerParameter

minor cleanup

add NetParameterPrettyPrint so that upgrade tool prints inputs before layers

add NetParameter required version number as breaking change for V0NetParameter

rollback previous commit adding version number to NetParameter -- going a different route

HDF5DataParameter message and concat_param

caffe.proto: layer->layers

add v0->v1 'bridge' proto and add util that uses it

move caffe.proto.v0 -> deprecated/caffe.v0.proto and add separate target makefile target for it

add duplicated params from InnerProductParam to ConvolutionParam and PoolingParam etc, create InfogainLossParam

NetParameter.layers -> layer

remove LayerConnection from proto, bottom and top now in LayerParameter

move LayerParameter and individual layer param messages to bottom of caffe.proto

move individual layer parameters to individual proto messages

fix caffe.proto style bugs

put inputs before layers in the proto so they print in that order

LayerType enum

fix layertype alphabetization

rebase and fix stuff, incorporate image and padding layers

incorporate WindowDataLayer

remove padding layer

update deprecated protos to latest dev versions

add V0NetParameter and UpgradeV0Net

2014-03-27

Standardize copyright, add root-level CONTRIBUTORS credit

2014-03-23

Implement and test HDF5OutputLayer

2014-03-20

set default to the best value

some cleanup

support for tightest square mode while finetuning

support for adding padding to windows in the window_data_layer

add window data layer

2014-03-18

fix style bugs in new layers' proto fields

Back-merge documentation and historical PRs to master

2014-02-28

Added concat_dim to caffe.proto and ConcatLayer to set of layers

2014-02-17

Fixed typos to pass test_images_layer

Renamed input_layer to images_layer

Fixed input_layer to pass tests, added cat image to data to perform the tests

Added the option to resize_image to resize images using cv::resize while reading them

2014-02-03

Update caffe.proto

Added device_id to solver.prototxt and to solver.cpp

2013-12-02

Several changes: changed the paths for the mnist demo removed unused fields in SolverParameter added solver_mode to specify CPU/GPU solving accordingly, removed the set mode code in the examples code removed examples/test_read_imagenet since it is trivial removed get_cifar script since no cifar demo exists.

2013-11-22

fix Makefile problem

2013-11-16

Made a major change: when initializing a network, the input size are no longer provided by an additional vector of blobs, but should be specified in the netparameters proto by the field "input_dim". This avoid the often awkward code of creating a dummy input vector just for the sake of initializing the network.

2013-11-08

misc update

2013-11-06

distributed server update. bug in synchronous connections.

data layer: random skip

2013-11-04

distributed solver - not checked thoroughly

2013-10-31

solver restructuring: now all prototxt are specified in the solver protocol buffer

2013-10-30

pooling layer: added stochastic pooling layer, not tested, compilable

2013-10-26

regarding https://plus.google.com/113952791760990667476/posts/Q5fwQY6JeEq - stopped the adagrad attempt.

2013-10-24

solver stepsize: float -> int32

2013-10-22

allow setting custom weight decay

2013-10-21

unchecked: using mean file.

2013-10-15

solver: added snapshotting ability

Moved the layer factory implementation to cpp; added snapshot and restore functions to solver.

2013-10-14

added custom learning rate for individual blobs

2013-10-10

misc update

blob math preparation

dropout serious bugfix. Seems to be working...

2013-10-07

misc update

2013-10-01

datalayer random cropping, not tested

2013-09-30

solver

data layer: allow scaling and subtraction

2013-09-29

a bunch of updates.

2013-09-27

updated a bunch of things, ready to test if it breaks things

started writing solver

2013-09-25

data layer modification, and net header

2013-09-24

renaming

a few updates

proto update

2013-09-23

copyright message

pooling layer cpu

2013-09-22

naming. I might regret it someday.

minor updates

2013-09-19

misc update

padding layer cuda code, need debug

2013-09-18

gemm util

updates

2013-09-17

misc update

misc update

2013-09-16

working halfway into dropout, machine down, changing machine

misc update

2013-09-14

misc update