doc update
doc update
doc update
added missing dummy header line in cascade descriptor: '#FeatureScalingLambda None'
Support for wrting out adaboost descriptor in the older cascade format as well
[docs] added manifest
[license] added TI copyright
Update LICENSE
Update LICENSE
[DOCS] Updated Markdown documentation
[EXTRACT] Added function for extraction.
[Vehichle, Trafficsign] Added support for more objects
[Vehichle, Trafficsign] Added support for more objects
[DATASET] Add support for training custom datasets
[detOffset] Added Detector parameter for detOffset
[CONFIG] Pyramid padding adjustment
[PATHS] Minor path changes in temporary data
[acfJacintoTrainTest] New interface for example jacinto script
[dbExtractList] Added a new, more flexible extraction script
[64x64] no feature approximation, 6 iterations, best results so far
[vbbLabeler] set maxCache to 0. Allows opening of variable size sequences.
[64x64] Good results after disabling feature approximnation
[32x64] Removed lambda scale approximation of features
[64x64] Best config with 64x64
[Bootstrap] Include positiv images also in bootstrap
[Freatures] Offset for gradient features
[Detection] Provided seperate detection threshold
[Freatures] Offset for gradient features
[Detection] Provided seperate detection threshold
[QualityAnalysis] In progress
[AccurateFastGgrad] Added and 'accurate' mode for fast gradient computation
[normRad] Added normRad for fast gradient
[PREPROC] Input pre-processing in progress
[FASTGRAD] Fast gradient computation
[FAST MODE] In progress
[BITMATCH] Analysis in progress
[DESCRIPTOR] Writing out a text descriptor
[Eval] Updated the Eval script to pass DB path.
[AcfJacinto32x64] Added cb folder from caltech site, initial training with smaller model
doc/: minor update to documentation
project page now lives at: https://pdollar.github.io/toolbox/
*VERSION 3.50 OF TOOLBOX*
channels/private/gradientMex.cpp: tiny fix (thanks Thabo Beeler)
classify/private/binaryTreeTrain1.cpp: minor fix that could potentially cause incorrect error computation (should not be triggered in practical cases). Thanks Tuan Dao
Merge pull request #2 from jltmtz/master
bbapply-embed boundary condition check (thanks Julieta!)
bbapply-embed boundary condition check (thanks Julieta!)
Update bbApply.m
acfTrain.m: minor fix for case when using cropped windows for training (thanks Jiangyu Liu)
videos/opticalFlow.m: minor fix to deal w degenerate patches
detector/acfSweeps.m: adding ACF parameter sweeps code
*VERSION 3.40 OF TOOLBOX*
*updated headers to reflect toolbox version 3.40*
detectors/acfReadme.m: added proper readme, updated demos to point to readme
detectors/models/: updated all models based on new settings
detector/acfTrain.m: proper param passing and commenting for LDCF, compute lambdas and filters on positive windows w/o jittering
detector/models/: adding LDCF models, removing detector output from git
detector/: support for LDCF (NIPS2014 paper)
detector/: replaced ACfCaltech model with AcfCaltech+ model described in NIPS14 paper, updated demo setup instructions and data location
detector/: acfTrain: nicer logging, acfTest: option to modify detector prior to applying to test set, acfDemo*: updated demos and model files (retrained on Linux)
classify/adaBoostApply.m, binaryTree*.m: set default max threads to 16 for OMP code
changed all line endings to LF
*VERSION 3.30 OF TOOLBOX*
*updated headers to reflect toolbox version 3.30*
standardizing c/cpp comments and removing tabs
external/toolboxUpdateHeader.m: generalizing to update c headers
updated toolbox name and headers
detector/bbApply.m>draw(): minor fix (thanks Michele Pratusevich)
test
Create README.md
added readed
added readed
added Mac binaries (mexmaci64)
toolboxCompile.m: disable OMP on Mac by default (OMP is tricky to setup on Mac)
see: www.mathworks.com/matlabcentral/newsreader/view_thread/334250
see: www.mathworks.com/matlabcentral/newsreader/view_thread/334250
fixing compile warnings on MAC
*VERSION 3.26 OF TOOLBOX*
*updated headers to reflect toolbox version 3.26*
detector/bbGt.m: minor fix to support old PASCAL 2007 format
matlab/fevalDistr.m: added ability to specify job priority!
bbGt>bbLoad(): support for ImageNet format
detector/bbGt.m: evalRes(): optimized
bbGt.m: minor fix (thanks David Hall)
Merge branch 'master' of vision.caltech.edu:pdollar-toolbox
matlab/fevalDistr.m: additional params to HPC cluster, removed ability to split job into multiple jobs
matlab/fevalDistr.m: removed search for stalled jobs on HPC cluster
channels/private/gradientMex.cpp: minor fix to tri-linear interpolation (didn't actually affect results, but it could, thanks Chulmin Yeum cyeum@purdue.edu)
overview.html: minor updated to webpage
*VERSION 3.25 OF TOOLBOX*
*updated headers to reflect toolbox version 3.25*
external/toolboxCompile.m: added info on how to disable OpenMp support, also that Matlab 2011b is necessary for toolbox
minor update (reference to PAMI14 paper and minor tweak to chnsScaling.m)
*VERSION 3.24 OF TOOLBOX*
*updated headers to reflect toolbox version 3.24*
external/history.txt: updated history.txt
classify/kmeans2.m: minor tweak to initialization (add randomness to initialization to break ties)
classify/forestTrain.m: support for learning with structured labels given 'discretize' function mapping structured to class labels
forestTrain.m: store best prediction at each node (and also optionally output best prediction per tree)
classify/pca.m: minor
matlab/subsampleMatrix.m: deprecated, removed from main toolbox
classify/pca.m: added robust SVD computation (bug fix for Matlab 2013a on Windows)
classify/pca.m: less aggressive subsampling (computers have more memory now)
classify/pca.m: cleanup
forestTrain.m: added 'twoing' splitting criteria
forestTrain.m: added entropy based splitting criteria
classify/private/forestFindThr.cpp: return 'gain' (improvement) instead of impurity
classify/private/forestFindThr.cpp: major cleanup (variable renaming, etc)
channels/imResample.m: minor (create copy if output should be identical to input)
channels/private/gradientMex.cpp: minor (support for nOrients=0)
channels/imResample.m: minor (create copy if output should be identical to input)
channels/private/convConst.cpp: minor cleanup
medianw.m: added support for multidimensional arrays
medianw.m: added fast weighted median