Added "guided backprop" option for deconv, fixed bugs, updated to latest caffe #5
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
batch norm statistics are not learnable parameters subject to solver updates, so they must be shielded from the solver. `BatchNorm` layer now masks its statistics for itself by zeroing parameter learning rates instead of relying on the layer definition. n.b. declaring `param`s for batch norm layers is no longer allowed.
automatically strip old batch norm layer definitions including `param`
messages. the batch norm layer used to require manually masking its
state from the solver by setting `param { lr_mult: 0 }` messages for
each of its statistics. this is now handled automatically by the layer.
[examples] Fixed typos in examples/cpp_classification/readme
Batch Norm: Further Documentation and Simplified Definition
fix layerSetUp of scale_layer to not add bias blob when already present
[TravisCI] google/protobuf renamed the 3.0 branch
Ignore Visual Studio Metadata
…_check slightly relax batch norm check
NV changed path to cuDNN
fixes the broken glog link in yum_install.md which is currently returning a 404.
fix typo in pascal_multilabel_datalayers.py
Fix: docs/yum_install.md glog broken link
Size checks are needed for loading parameters to avoid strange bugs when loading data we continue to reshape.
Sigmoid Cross-Entropy Loss: add GPU forward for full GPU mode speed-up
a trival commit which adds the missing star ;)
add the missing star in comment
corrected typo in accuracy_layer.hpp: MaxTopBlos -> MaxTopBlobs
For information on Github issue templates, see: https://github.com/blog/2111-issue-and-pull-request-templates The template has been revised according to discussion with @shelhamer and @willyd on pull request BVLC#4914.
Add GitHub issue template to curb misuse
Handling destruction of empty Net objects
Rewrite crop layer GPU implementation
Downgrade boost requirement from 1.55 to 1.54
docs/debian guide: update compiler combination table
…o-library cmake: rename libproto.a -> libcaffeproto.a
List branches in readme
Caffe package is available for APT since Debian 9.0 and Ubuntu 17.04 .
Fixed typo "locaitons " -> "location" in line 17
Version 0.18 moved cross-validation to sklearn.model_selection - see http://scikit-learn.org/stable/whats_new.html#version-0-18 Version 0.17 deprecated class_weight="auto" in favor of class_weight="balanced"
pd.scatter_matrix -> pd.plotting.scatter_matrix
API updates in Logistic Regression example
[docs] fix spelling for LRN
[docs] update apt installation guide for Debian and Ubuntu
User may not install xcoder into default directory especially if there are two different versions of xcoder installed.
[cmake] fix hardcode xcode path
Mention SKX support
Fix division operator for Compatibility of python 3 in classifier.py
The pointers could be used by CUDA wrapper libraries in Python such as PyCUDA, gnumpy, Theano etc.
Expose GPU pointers to Python
… 'Zeiler & Fergus' - fix bug related to layer name vs blob name confusion
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I've added another option for making the deconv, namely using "guided backpropagation" instead of "Zeiler-Fergus" deconv. Note that the type of deconv to use is passed as a parameter so one can see trigger the ZF backprop.
In addition I've fixed some bugs, mainly around the confusion between the blob name and layer name.
Finally, this I've rebased to use the latest caffe.
This version is needed for the updated version of deep-visualization-toolbox