remove linking against mkl_intel_thread: unneeded
and gives hard-to-debug errors on os x
and gives hard-to-debug errors on os x
Merge pull request #62 from viirya/master
Fix example path + set number of CUDA threads according to architecture
Fix example path + set number of CUDA threads according to architecture
Fix test_data_layer segfault by adding destructor to join pthread
note pretrained model licensing: academic / non-commercial use only
read single input, load/save csv, and record windows
Load `--input_file` as a list of image filenames if .txt OR read as CSV
with fields filename,ymin,xmin,ymax,xmax (with labeled header) if .csv.
Save `--output_file` as HDF5 if .h5 or CSV if .csv. For CSV, enumerate
the class probabilities as numbered class fields.
Record crop windows in original image coordinates if `center_only` or
`corners` crop mode selected. Previously, these modes didn't report
locations.
Load `--input_file` as a list of image filenames if .txt OR read as CSV
with fields filename,ymin,xmin,ymax,xmax (with labeled header) if .csv.
Save `--output_file` as HDF5 if .h5 or CSV if .csv. For CSV, enumerate
the class probabilities as numbered class fields.
Record crop windows in original image coordinates if `center_only` or
`corners` crop mode selected. Previously, these modes didn't report
locations.
remove outdated reference to input size, debugging print
choose proper thread number per block according to CUDA architecture.
fix program path.
align detector comments to reality
Merge pull request #61 from kloudkl/master
Save the last batch of data in image set conversion
+ comment and output edits
Save the last batch of data in image set conversion
+ comment and output edits
Save the last batch of data in image set conversion
detect by window list
Do window detection on a list of predefined windows by adding a
crop_mode `list` that accepts a path to a text file of format:
image_filename xmin ymin xmax ymax
Do window detection on a list of predefined windows by adding a
crop_mode `list` that accepts a path to a text file of format:
image_filename xmin ymin xmax ymax
replace magic numbers with variable names in comments
fix ImageNet solver max iteration typo
max iteration no. is 450,000 (= 90 epochs)
caught by @SWu
https://github.com/BVLC/caffe/pull/33#issuecomment-33175810
max iteration no. is 450,000 (= 90 epochs)
caught by @SWu
https://github.com/BVLC/caffe/pull/33#issuecomment-33175810
automagically set detection batch size from network
document power_wrapper -> detector in detection notebook
Merge pull request #56 from shelhamer/detection-wrapper
Make detection module for python wrapper
Make detection module for python wrapper
promote power_wrapper to 'detection' submodule
groom power_wrapper flags (cont'd)
note power_wrapper TODOs
generalize power_wrapper to different networks and inputs
generalize from the imagenet reference to other detection networks:
- refactor to configure for a given net and data source
- determine dimensions of net input and output automatically from blobs
generalize from the imagenet reference to other detection networks:
- refactor to configure for a given net and data source
- determine dimensions of net input and output automatically from blobs
groom power_wrapper flags
python lint
ignore distribute dir
give batch size efficiency advice
default power_wrapper batch size to 10 (aeca741a69 cont'd)
default command line arg and function arg to 10
remove global BATCH_SIZE in favor of arg
default command line arg and function arg to 10
remove global BATCH_SIZE in favor of arg
Merge pull request #51 from sguada/solver_test
Do snapshot after computing loss and test accuracy
Do snapshot after computing loss and test accuracy
Do snapshot after computing loss and test accuracy
include install notes
include intro, license, and citing in README
License under BSD
Caffe is henceforth BSD licensed and welcomes community contributions!
Caffe is henceforth BSD licensed and welcomes community contributions!
Merge pull request #11 from longjon/master
Python interface to blobs and blob data through boost::python
Python interface to blobs and blob data through boost::python
cleanup whitespace
Merge pull request #29 from kloudkl/master
Separate build, distribute, and src directories + build more cuda archs
Separate build, distribute, and src directories + build more cuda archs
Merge pull request #34 from forresti/master
set default power_wrapper batch size to match the imagenet_deploy
compatible out-of-the-box with the ipython notebook tutorial
set default power_wrapper batch size to match the imagenet_deploy
compatible out-of-the-box with the ipython notebook tutorial
setting default power_wrapper batch size to match the imagenet_deploy prototxt
Expose params in Python interface
Add Python interface to layer blobs
Add removing distribute directory when make clean
Add make proto target to seperately generate caffe/proto/caffe.pb.h
issue: #22
issue: #22
Seperated build, distribute and source directories
issue: #1
issue: #1
Merge pull request #18 from jeffdonahue/bvlc
change imagenet_val batch size from 200 to intended 50
change imagenet_val batch size from 200 to intended 50
change imagenet_val batch size from 200 to intended 50
point caffe url to bvlc
bringing license up to date with gh-pages version
updated detection demo notebook with picture of two cats
Merge pull request #10 from sergeyk/selective_search_proposals2
Efficient processing of images, with option to use selective search window proposals
Efficient processing of images, with option to use selective search window proposals
minor edit
selective search notebook and renaming to power_wrapper
fixed bug that renormalized window crops on second resize
center_only and corners modes work correctly
processing images in batch, with option to use selective search window proposals
renaming"
Update README.md
Merge pull request #9 from jeffdonahue/fixflattenbug
fix really stupid bug in flatten layer (and add test that shows the
fix really stupid bug in flatten layer (and add test that shows the
fix really stupid bug in flatten layer (and add test that shows the
failure case; not sure why CheckGradientExhaustive didn't catch it)
failure case; not sure why CheckGradientExhaustive didn't catch it)
lenet.prototxt loss layer rename
mnist train script
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.
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.
get_mnist.sh: changed the script to generate leveldb as well.
convert script: spacing
Merge pull request #7 from jeffdonahue/headers
modify makefile to recompile on changes to header files
modify makefile to recompile on changes to header files
modify makefile to recompile on changes to header files
Update README.md
Merge pull request #6 from jeffdonahue/flatten
add flatten layer
add flatten layer
add flatten layer
net.cpp: LOG to DLOG
io.cpp: changed back CV_LOAD_IMAGE_COLOR for earlier versions of opencv
Update io.cpp
git ignore
net speed benchmark: set device 0
removed the -Wl flag which was an earlier experimental try
makefile update
Merge branch 'master' of github.com:Yangqing/caffe
makefile: add both lib and lib64 to the cuda library path.
makefile: added a MATLAB_DIR variable. pycaffe and matcaffe will not be compiled in default unless one calls make pycaffe or make matcaffe explicitly.
update Makefile and add some more docs
Merge remote-tracking branch 'origin/master'
cleanup and include ILSVRC mean image
first pass at matlab wrapper (somewhat messy still)
fix Makefile problem
fixed minor typo in imagenet wrapper
cleaned makefile a little bit
Merge branch 'master' of github.com:Yangqing/caffe
remove remaining distributed solver stuff
remove caffe header for distributed solver
removed the unsuccessful distributed solver code
fix pycaffe dependency
fixed the missing CV flag
makefile: not compile test in default
Merge branch 'master' of github.com:Yangqing/caffe
makefile: fixed the nvcc path
io.cpp: changed the imread flag to cv::IMREAD_COLOR
Merge branch 'master' of github.com:Yangqing/caffe
added explanation notes to syncedmem
imagenet_solver.prototxt: added back the missing test line, not sure when it slipped off the repo
Merge branch 'master' of github.com:Yangqing/caffe
imagenet deploy prototxt: instead of having a data layer, this network proto takes an external blob called data. The shape is hardcoded to accompany the imagenet python wrapper at python/caffe/imagenet/wrapper.py. You need to change the shape if you intend to use other shapes.
print status message on error opening leveldb
Merge branch 'master', remote-tracking branch 'origin'