Kai Li [Sun, 12 Jan 2014 07:56:47 +0000 (15:56 +0800)]
Fix test_data_layer segfault by adding destructor to join pthread
Evan Shelhamer [Fri, 31 Jan 2014 10:52:29 +0000 (02:52 -0800)]
note pretrained model licensing: academic / non-commercial use only
Evan Shelhamer [Fri, 31 Jan 2014 07:12:41 +0000 (23:12 -0800)]
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.
Evan Shelhamer [Fri, 31 Jan 2014 07:26:40 +0000 (23:26 -0800)]
remove outdated reference to input size, debugging print
Evan Shelhamer [Wed, 29 Jan 2014 02:09:13 +0000 (18:09 -0800)]
align detector comments to reality
Evan Shelhamer [Tue, 28 Jan 2014 09:59:27 +0000 (01:59 -0800)]
Merge pull request #61 from kloudkl/master
Save the last batch of data in image set conversion
+ comment and output edits
Kai Li [Tue, 28 Jan 2014 07:35:56 +0000 (15:35 +0800)]
Save the last batch of data in image set conversion
Evan Shelhamer [Mon, 27 Jan 2014 23:20:29 +0000 (15:20 -0800)]
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
Evan Shelhamer [Mon, 27 Jan 2014 23:17:45 +0000 (15:17 -0800)]
replace magic numbers with variable names in comments
Evan Shelhamer [Sun, 26 Jan 2014 06:33:18 +0000 (22:33 -0800)]
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
Evan Shelhamer [Sun, 26 Jan 2014 04:53:00 +0000 (20:53 -0800)]
automagically set detection batch size from network
Evan Shelhamer [Sun, 26 Jan 2014 03:58:18 +0000 (19:58 -0800)]
document power_wrapper -> detector in detection notebook
Jeff Donahue [Sun, 26 Jan 2014 02:04:56 +0000 (18:04 -0800)]
Merge pull request #56 from shelhamer/detection-wrapper
Make detection module for python wrapper
Evan Shelhamer [Sat, 25 Jan 2014 04:51:00 +0000 (20:51 -0800)]
promote power_wrapper to 'detection' submodule
Evan Shelhamer [Sat, 25 Jan 2014 04:33:24 +0000 (20:33 -0800)]
groom power_wrapper flags (cont'd)
Evan Shelhamer [Sat, 25 Jan 2014 03:31:25 +0000 (19:31 -0800)]
note power_wrapper TODOs
Evan Shelhamer [Sat, 25 Jan 2014 03:26:31 +0000 (19:26 -0800)]
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
Evan Shelhamer [Sat, 25 Jan 2014 03:24:58 +0000 (19:24 -0800)]
groom power_wrapper flags
Evan Shelhamer [Sat, 25 Jan 2014 01:36:46 +0000 (17:36 -0800)]
python lint
Evan Shelhamer [Sat, 25 Jan 2014 00:18:17 +0000 (16:18 -0800)]
ignore distribute dir
Evan Shelhamer [Fri, 24 Jan 2014 05:15:26 +0000 (21:15 -0800)]
give batch size efficiency advice
Evan Shelhamer [Fri, 24 Jan 2014 05:05:17 +0000 (21:05 -0800)]
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
Yangqing Jia [Fri, 24 Jan 2014 02:09:43 +0000 (18:09 -0800)]
Merge pull request #51 from sguada/solver_test
Do snapshot after computing loss and test accuracy
Sergio Guadarrama [Wed, 22 Jan 2014 23:13:22 +0000 (15:13 -0800)]
Do snapshot after computing loss and test accuracy
Evan Shelhamer [Wed, 22 Jan 2014 02:50:07 +0000 (18:50 -0800)]
include install notes
Evan Shelhamer [Wed, 22 Jan 2014 02:37:51 +0000 (18:37 -0800)]
include intro, license, and citing in README
Evan Shelhamer [Tue, 21 Jan 2014 02:24:41 +0000 (18:24 -0800)]
License under BSD
Caffe is henceforth BSD licensed and welcomes community contributions!
Evan Shelhamer [Mon, 20 Jan 2014 05:17:43 +0000 (21:17 -0800)]
Merge pull request #11 from longjon/master
Python interface to blobs and blob data through boost::python
Evan Shelhamer [Sun, 19 Jan 2014 22:34:12 +0000 (14:34 -0800)]
cleanup whitespace
Evan Shelhamer [Sun, 19 Jan 2014 22:10:14 +0000 (14:10 -0800)]
Merge pull request #29 from kloudkl/master
Separate build, distribute, and src directories + build more cuda archs
Evan Shelhamer [Sun, 19 Jan 2014 06:59:33 +0000 (22:59 -0800)]
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
forresti [Wed, 15 Jan 2014 20:02:12 +0000 (12:02 -0800)]
setting default power_wrapper batch size to match the imagenet_deploy prototxt
Jonathan L Long [Tue, 14 Jan 2014 23:44:53 +0000 (15:44 -0800)]
Expose params in Python interface
Jonathan L Long [Sat, 7 Dec 2013 00:39:01 +0000 (16:39 -0800)]
Add Python interface to layer blobs
Kai Li [Mon, 13 Jan 2014 12:42:01 +0000 (20:42 +0800)]
Add removing distribute directory when make clean
Kai Li [Mon, 13 Jan 2014 12:25:30 +0000 (20:25 +0800)]
Add make proto target to seperately generate caffe/proto/caffe.pb.h
issue: #22
Kai Li [Mon, 13 Jan 2014 11:34:34 +0000 (19:34 +0800)]
Seperated build, distribute and source directories
issue: #1
Jeff Donahue [Sun, 22 Dec 2013 21:24:34 +0000 (13:24 -0800)]
Merge pull request #18 from jeffdonahue/bvlc
change imagenet_val batch size from 200 to intended 50
Jeff Donahue [Sun, 22 Dec 2013 21:13:47 +0000 (13:13 -0800)]
change imagenet_val batch size from 200 to intended 50
Evan Shelhamer [Wed, 18 Dec 2013 03:21:15 +0000 (19:21 -0800)]
point caffe url to bvlc
Sergey Karayev [Fri, 6 Dec 2013 20:39:49 +0000 (12:39 -0800)]
bringing license up to date with gh-pages version
Sergey Karayev [Fri, 6 Dec 2013 19:28:41 +0000 (11:28 -0800)]
updated detection demo notebook with picture of two cats
Yangqing Jia [Fri, 6 Dec 2013 04:14:15 +0000 (20:14 -0800)]
Merge pull request #10 from sergeyk/selective_search_proposals2
Efficient processing of images, with option to use selective search window proposals
Sergey Karayev [Fri, 6 Dec 2013 03:35:32 +0000 (19:35 -0800)]
minor edit
Sergey Karayev [Fri, 6 Dec 2013 03:33:13 +0000 (19:33 -0800)]
selective search notebook and renaming to power_wrapper
Sergey Karayev [Fri, 6 Dec 2013 03:12:50 +0000 (19:12 -0800)]
fixed bug that renormalized window crops on second resize
Sergey Karayev [Fri, 6 Dec 2013 01:21:47 +0000 (17:21 -0800)]
center_only and corners modes work correctly
Sergey Karayev [Fri, 6 Dec 2013 00:38:36 +0000 (16:38 -0800)]
processing images in batch, with option to use selective search window proposals
Yangqing Jia [Thu, 5 Dec 2013 17:38:11 +0000 (09:38 -0800)]
renaming"
Yangqing Jia [Thu, 5 Dec 2013 09:13:59 +0000 (01:13 -0800)]
Update README.md
Yangqing Jia [Thu, 5 Dec 2013 04:24:26 +0000 (20:24 -0800)]
Merge pull request #9 from jeffdonahue/fixflattenbug
fix really stupid bug in flatten layer (and add test that shows the
Jeff Donahue [Thu, 5 Dec 2013 04:16:25 +0000 (20:16 -0800)]
fix really stupid bug in flatten layer (and add test that shows the
failure case; not sure why CheckGradientExhaustive didn't catch it)
Yangqing Jia [Tue, 3 Dec 2013 18:53:29 +0000 (10:53 -0800)]
lenet.prototxt loss layer rename
Yangqing Jia [Tue, 3 Dec 2013 18:05:19 +0000 (10:05 -0800)]
mnist train script
Yangqing Jia [Mon, 2 Dec 2013 23:44:43 +0000 (15:44 -0800)]
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.
Yangqing Jia [Mon, 2 Dec 2013 23:19:37 +0000 (15:19 -0800)]
get_mnist.sh: changed the script to generate leveldb as well.
Yangqing Jia [Mon, 2 Dec 2013 00:38:24 +0000 (16:38 -0800)]
convert script: spacing
Yangqing Jia [Thu, 28 Nov 2013 22:51:16 +0000 (14:51 -0800)]
Merge pull request #7 from jeffdonahue/headers
modify makefile to recompile on changes to header files
Jeff Donahue [Thu, 28 Nov 2013 21:55:17 +0000 (13:55 -0800)]
modify makefile to recompile on changes to header files
Yangqing Jia [Thu, 28 Nov 2013 18:42:30 +0000 (10:42 -0800)]
Update README.md
Yangqing Jia [Thu, 28 Nov 2013 05:39:08 +0000 (21:39 -0800)]
Merge pull request #6 from jeffdonahue/flatten
add flatten layer
Jeff Donahue [Thu, 28 Nov 2013 05:29:05 +0000 (21:29 -0800)]
add flatten layer
Yangqing Jia [Thu, 28 Nov 2013 04:54:21 +0000 (20:54 -0800)]
net.cpp: LOG to DLOG
Yangqing Jia [Thu, 28 Nov 2013 04:54:01 +0000 (20:54 -0800)]
io.cpp: changed back CV_LOAD_IMAGE_COLOR for earlier versions of opencv
Yangqing Jia [Thu, 28 Nov 2013 01:00:00 +0000 (17:00 -0800)]
Update io.cpp
Yangqing Jia [Tue, 26 Nov 2013 21:16:12 +0000 (13:16 -0800)]
git ignore
Yangqing Jia [Tue, 26 Nov 2013 18:59:06 +0000 (10:59 -0800)]
net speed benchmark: set device 0
Yangqing Jia [Tue, 26 Nov 2013 17:49:29 +0000 (09:49 -0800)]
removed the -Wl flag which was an earlier experimental try
Yangqing Jia [Tue, 26 Nov 2013 17:47:26 +0000 (09:47 -0800)]
makefile update
Yangqing Jia [Mon, 25 Nov 2013 06:13:38 +0000 (22:13 -0800)]
Merge branch 'master' of github.com:Yangqing/caffe
Yangqing Jia [Mon, 25 Nov 2013 06:13:34 +0000 (22:13 -0800)]
makefile: add both lib and lib64 to the cuda library path.
Yangqing Jia [Sat, 23 Nov 2013 17:37:27 +0000 (09:37 -0800)]
makefile: added a MATLAB_DIR variable. pycaffe and matcaffe will not be compiled in default unless one calls make pycaffe or make matcaffe explicitly.
Ross Girshick [Fri, 22 Nov 2013 22:16:58 +0000 (14:16 -0800)]
update Makefile and add some more docs
Ross Girshick [Fri, 22 Nov 2013 22:03:55 +0000 (14:03 -0800)]
Merge remote-tracking branch 'origin/master'
Ross Girshick [Fri, 22 Nov 2013 22:03:36 +0000 (14:03 -0800)]
cleanup and include ILSVRC mean image
Ross Girshick [Fri, 22 Nov 2013 21:16:23 +0000 (13:16 -0800)]
first pass at matlab wrapper (somewhat messy still)
Yangqing Jia [Fri, 22 Nov 2013 01:00:57 +0000 (17:00 -0800)]
fix Makefile problem
Ross Girshick [Thu, 21 Nov 2013 23:45:04 +0000 (15:45 -0800)]
fixed minor typo in imagenet wrapper
Yangqing Jia [Thu, 21 Nov 2013 21:31:18 +0000 (13:31 -0800)]
cleaned makefile a little bit
Yangqing Jia [Thu, 21 Nov 2013 19:19:12 +0000 (11:19 -0800)]
Merge branch 'master' of github.com:Yangqing/caffe
Yangqing Jia [Thu, 21 Nov 2013 19:16:08 +0000 (11:16 -0800)]
remove remaining distributed solver stuff
Yangqing Jia [Thu, 21 Nov 2013 19:16:08 +0000 (11:16 -0800)]
remove caffe header for distributed solver
Yangqing Jia [Thu, 21 Nov 2013 19:11:46 +0000 (11:11 -0800)]
removed the unsuccessful distributed solver code
Yangqing Jia [Tue, 19 Nov 2013 23:05:39 +0000 (15:05 -0800)]
fix pycaffe dependency
Yangqing Jia [Tue, 19 Nov 2013 22:37:19 +0000 (14:37 -0800)]
fixed the missing CV flag
Yangqing Jia [Tue, 19 Nov 2013 22:31:42 +0000 (14:31 -0800)]
makefile: not compile test in default
Yangqing Jia [Tue, 19 Nov 2013 22:03:56 +0000 (14:03 -0800)]
Merge branch 'master' of github.com:Yangqing/caffe
Yangqing Jia [Tue, 19 Nov 2013 21:39:20 +0000 (13:39 -0800)]
makefile: fixed the nvcc path
Yangqing Jia [Tue, 19 Nov 2013 21:37:28 +0000 (13:37 -0800)]
io.cpp: changed the imread flag to cv::IMREAD_COLOR
Yangqing Jia [Tue, 19 Nov 2013 19:06:12 +0000 (11:06 -0800)]
Merge branch 'master' of github.com:Yangqing/caffe
Yangqing Jia [Tue, 19 Nov 2013 17:08:31 +0000 (09:08 -0800)]
added explanation notes to syncedmem
Yangqing Jia [Tue, 19 Nov 2013 16:51:43 +0000 (08:51 -0800)]
imagenet_solver.prototxt: added back the missing test line, not sure when it slipped off the repo
Yangqing Jia [Tue, 19 Nov 2013 00:06:37 +0000 (16:06 -0800)]
Merge branch 'master' of github.com:Yangqing/caffe
Yangqing Jia [Tue, 19 Nov 2013 00:06:33 +0000 (16:06 -0800)]
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.
Jeff Donahue [Sun, 17 Nov 2013 22:39:42 +0000 (14:39 -0800)]
print status message on error opening leveldb
Jeff Donahue [Sun, 17 Nov 2013 21:45:22 +0000 (13:45 -0800)]
Merge branch 'master', remote-tracking branch 'origin'
Yangqing Jia [Sun, 17 Nov 2013 02:41:28 +0000 (18:41 -0800)]
removed the StillFresh function that was used to make sure training does not happen with an earlier version of code.
Yangqing Jia [Sat, 16 Nov 2013 00:58:53 +0000 (16:58 -0800)]
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.
Yangqing Jia [Fri, 15 Nov 2013 16:02:48 +0000 (08:02 -0800)]
wrapper update
Yangqing Jia [Fri, 15 Nov 2013 05:12:23 +0000 (21:12 -0800)]
finetune code