diff --git a/Code/FeatExtraction/ClassifMonoView.py b/Code/FeatExtraction/ClassifMonoView.py
index c3797c43ff4151f0025490ca5977049928a1d2ac..fc7fa2d171319ae941cd15a82df1be0e9caf64aa 100644
--- a/Code/FeatExtraction/ClassifMonoView.py
+++ b/Code/FeatExtraction/ClassifMonoView.py
@@ -135,7 +135,7 @@ def calcTrainTest(X,y,split):
# X_test: Test Data
# y_test: Test Labels
# num_estimators: number of trees
-def calcClassifRandomForestCV(X_train, y_train, num_estimators):
+def calcClassifRandomForestCV(X_train, y_train, num_estimators, cv_folds, clas_cores):
# PipeLine with RandomForest classifier
pipeline_rf = Pipeline([('classifier', RandomForestClassifier())])
@@ -143,7 +143,6 @@ def calcClassifRandomForestCV(X_train, y_train, num_estimators):
# can be extended with: oob_score, min_samples_leaf, max_features
param_rf = { 'classifier__n_estimators': num_estimators}
- kfolds = 5
# pipeline: Gridsearch avec le pipeline comme estimator
# param: pour obtenir le meilleur model il va essayer tous les possiblites
# refit: pour utiliser le meilleur model apres girdsearch
@@ -154,9 +153,9 @@ def calcClassifRandomForestCV(X_train, y_train, num_estimators):
pipeline_rf,
param_grid=param_rf,
refit=True,
- n_jobs=1,
+ n_jobs=clas_cores,
scoring='accuracy',
- cv=kfolds,
+ cv=cv_folds,
)
rf_detector = grid_rf.fit(X_train, y_train)
@@ -167,26 +166,26 @@ def calcClassifRandomForestCV(X_train, y_train, num_estimators):
return (description, rf_detector)
-def calcClassifRandomForest(X_train, X_test, y_test, y_train, num_estimators):
- from sklearn.grid_search import ParameterGrid
- param_rf = { 'classifier__n_estimators': num_estimators}
- forest = RandomForestClassifier()
-
- bestgrid=0;
- for g in ParameterGrid(grid):
- forest.set_params(**g)
- forest.fit(X_train,y_train)
- score = forest.score(X_test, y_test)
-
- if score > best_score:
- best_score = score
- best_grid = g
-
- rf_detector = RandomForestClassifier()
- rf_detector.set_params(**best_grid)
- rf_detector.fit(X_train,y_train)
+#def calcClassifRandomForest(X_train, X_test, y_test, y_train, num_estimators):
+# from sklearn.grid_search import ParameterGrid
+# param_rf = { 'classifier__n_estimators': num_estimators}
+# forest = RandomForestClassifier()
+#
+# bestgrid=0;
+# for g in ParameterGrid(grid):
+# forest.set_params(**g)
+# forest.fit(X_train,y_train)
+# score = forest.score(X_test, y_test)
+#
+# if score > best_score:
+# best_score = score
+# best_grid = g
+#
+# rf_detector = RandomForestClassifier()
+# rf_detector.set_params(**best_grid)
+# rf_detector.fit(X_train,y_train)
- #desc_estimators = best_grid
- description = "Classif_" + "RF" + "-" + "CV_" + "NO" + "-" + "Trees_" + str(best_grid)
+# #desc_estimators = best_grid
+# description = "Classif_" + "RF" + "-" + "CV_" + "NO" + "-" + "Trees_" + str(best_grid)
- return (description, rf_detector)
\ No newline at end of file
+# return (description, rf_detector)
\ No newline at end of file
diff --git a/Code/FeatExtraction/DBCrawl.py b/Code/FeatExtraction/DBCrawl.py
index fb6078eac4ef98e597d7fddab72ed25ba60a86fc..1312d77a771e7fcd9660ae399d1af114be077416 100644
--- a/Code/FeatExtraction/DBCrawl.py
+++ b/Code/FeatExtraction/DBCrawl.py
@@ -4,9 +4,9 @@
# Import built-in modules
import os # for iteration throug directories
-import pandas as pd # for Series and DataFrames
# Import 3rd party modules
+import pandas as pd # for Series and DataFrames
# Import own modules
diff --git a/Code/FeatExtraction/ExecClassifMV.py b/Code/FeatExtraction/ExecClassifMonoView.py
similarity index 55%
rename from Code/FeatExtraction/ExecClassifMV.py
rename to Code/FeatExtraction/ExecClassifMonoView.py
index 51663e130a492715e97f75c5581a849883380c4e..692caad1131c964c91fb1837fb9e6b71a2c49a86 100644
--- a/Code/FeatExtraction/ExecClassifMV.py
+++ b/Code/FeatExtraction/ExecClassifMonoView.py
@@ -3,20 +3,12 @@
""" Script to perform feature parameter optimisation """
# Import built-in modules
-#import cv2 # for OpenCV
-#import cv # for OpenCV
-#import datetime # for TimeStamp in CSVFile
-#from scipy.cluster.vq import * # for Clustering http://docs.scipy.org/doc/scipy/reference/cluster.vq.html
-#import numpy as np # for arrays
-#import time # for time calculations
from argparse import ArgumentParser # for acommand line arguments
# Import 3rd party modules
# Import own modules
-#import DBCrawl # Functions to read Images from Database
-#import ExportResults # Functions to render results
-#import FeatExtraction # Functions to extract the features from Database
+
# Author-Info
__author__ = "Nikolas Huelsmann"
diff --git a/Code/FeatExtraction/ExecFeatExtraction.py b/Code/FeatExtraction/ExecFeatExtraction.py
index 5ac151b3681f29c2f5f6b8e2b5482f963fbb05df..7412ce13043b580c353cb38b9f02e3cd679641e5 100644
--- a/Code/FeatExtraction/ExecFeatExtraction.py
+++ b/Code/FeatExtraction/ExecFeatExtraction.py
@@ -3,13 +3,10 @@
""" Script to perform feature parameter optimisation """
# Import built-in modules
-import cv2 # for OpenCV
-import cv # for OpenCV
-import datetime # for TimeStamp in CSVFile
-from scipy.cluster.vq import * # for Clustering http://docs.scipy.org/doc/scipy/reference/cluster.vq.html
-import numpy as np # for arrays
-import time # for time calculations
-from argparse import ArgumentParser # for acommand line arguments
+import datetime # for TimeStamp in CSVFile
+import time # for time calculations
+import argparse # for acommand line arguments
+import textwrap
# Import 3rd party modules
@@ -25,52 +22,53 @@ __date__ = 2016-03-10
### Argument Parser
-parser = ArgumentParser(description='Export Features')
+parser = argparse.ArgumentParser(
+description='This methods permits to export one or more features at the same time for a database of images (path, name). To extract one feature activate it by using the specific argument (e.g. -RGB). For each feature you can define the parameters by using the optional arguments (e.g. --RGB_Hist 32). The results will be exported to a CSV-File.',
+formatter_class=argparse.ArgumentDefaultsHelpFormatter)
-parser.add_argument('--name', action='store', help='Name of DB, default DB', default='DB')
-parser.add_argument('--path', action='store', help='Path to the database e.g. D:\\Caltech', default='D:\\CaltechMini')
-parser.add_argument('--cores', action='store', help='Number of cores used for parallelization of HOG, default 1', type=int, default=1)
+groupStandard = parser.add_argument_group('necessary arguments:')
+groupStandard.add_argument('--name', metavar='STRING', action='store', help='Select a name of DB, e.g. Caltech (default: %(default)s)', default='DB')
+groupStandard.add_argument('--path', metavar='STRING', action='store', help='Path to the database (default: %(default)s)', default='D:\\CaltechMini')
-parser.add_argument('--RGB', action='store_true', help='Use option to activate RGB')
-parser.add_argument('--RGB_Hist', action='store', help='RGB: Number of bins for histogram, default 16', type=int, default=16)
-parser.add_argument('--RGB_CI', action='store', help='RGB: Max Color Intensity [0 to VALUE], default 256', type=int, default=256)
-parser.add_argument('--RGB_NMinMax', action='store_true', help='RGB: Use option to actvate MinMax Norm, default distribtion')
-parser.add_argument('--HSV', action='store_true', help='Use option to activate HSV')
-parser.add_argument('--HSV_H', action='store', help='HSV: Number of bins for Hue, default 8', type=int, default=8)
-parser.add_argument('--HSV_S', action='store', help='HSV: Number of bins for Saturation, default 3', type=int, default=3)
-parser.add_argument('--HSV_V', action='store', help='HSV: Number of bins for Value, default 3', type=int, default=3)
-parser.add_argument('--HSV_NMinMax', action='store_true', help='HSV: Use option to actvate MinMax Norm, default distribtion')
-
-parser.add_argument('--SIFT', action='store_true', help='Use option to activate SIFT')
-parser.add_argument('--SIFT_Cluster', action='store', help='SIFT: Number of k-means cluster, default 50', type=int, default=50)
-parser.add_argument('--SIFT_NMinMax', action='store_true', help='SIFT: Use option to actvate MinMax Norm, default distribtion')
-
-parser.add_argument('--SURF', action='store_true', help='Use option to activate SURF')
-parser.add_argument('--SURF_Cluster', action='store', help='SURF: Number of k-means cluster, default 50', type=int, default=50)
-parser.add_argument('--SURF_NMinMax', action='store_true', help='SURF: Use option to actvate MinMax Norm, default distribtion')
-
-parser.add_argument('--HOG', action='store_true', help='Use option to activate HOG')
-parser.add_argument('--HOG_CellD', action='store', help='HOG: CellDimension for local histograms, default 5', type=int, default=5)
-parser.add_argument('--HOG_Orient', action='store', help='HOG: Number of bins of local histograms , default 8', type=int, default=8)
-parser.add_argument('--HOG_Cluster', action='store', help='HOG: Number of k-means cluster, default 12', type=int, default=12)
-parser.add_argument('--HOG_Iter', action='store', help='HOG: Max. number of iterations for clustering, default 100', type=int, default=100)
+groupRGB = parser.add_argument_group('RGB arguments:')
+groupRGB.add_argument('-RGB', action='store_true', help='Use option to activate RGB')
+groupRGB.add_argument('--RGB_Bins', metavar='INT', action='store', help='Number of bins for histogram', type=int, default=16)
+groupRGB.add_argument('--RGB_CI', metavar='INT', action='store', help='Max Color Intensity [0 to VALUE]', type=int, default=256)
+groupRGB.add_argument('-RGB_NMinMax', action='store_true', help='Use option to actvate MinMax Norm instead of Distribution')
+groupHSV = parser.add_argument_group('HSV arguments:')
+groupHSV.add_argument('-HSV', action='store_true', help='Use option to activate HSV')
+groupHSV.add_argument('--HSV_H_Bins', metavar='INT', action='store', help='Number of bins for Hue', type=int, default=16)
+groupHSV.add_argument('--HSV_S_Bins', metavar='INT', action='store', help='Number of bins for Saturation', type=int, default=4)
+groupHSV.add_argument('--HSV_V_Bins', metavar='INT', action='store', help='Number of bins for Value', type=int, default=4)
+groupHSV.add_argument('-HSV_NMinMax', action='store_true', help='Use option to actvate MinMax Norm instead of Distribution')
-# CELL_DIMENSION is the dimension of the cells on which we will compute local histograms
-# NB_ORIENTATIONS is the number of bins of this local histograms
-# intuitively, if CELL_DIMENSION is small it's better to have a small NB_ORIENTATIONS in order to have meaningful local histograms
-# NB_CLUSTERS is the number of bins of the global histograms (the number of clusters in the KMEANS algorithm used for the bag of word)
-# MAXITER is the maximum number of iteration for the clustering algorithm
-
+groupSIFT = parser.add_argument_group('SIFT arguments:')
+groupSIFT.add_argument('-SIFT', action='store_true', help='Use option to activate SIFT')
+groupSIFT.add_argument('--SIFT_Cluster', metavar='INT', action='store', help='Number of k-means cluster', type=int, default=50)
+groupSIFT.add_argument('-SIFT_NMinMax', action='store_true', help='Use option to actvate MinMax Norm instead of Distribution')
+
+groupSURF = parser.add_argument_group('SURF arguments:')
+groupSURF.add_argument('-SURF', action='store_true', help='Use option to activate SURF')
+groupSURF.add_argument('--SURF_Cluster', metavar='INT', action='store', help='Number of k-means cluster', type=int, default=50)
+groupSURF.add_argument('-SURF_NMinMax', action='store_true', help='Use option to actvate MinMax Norm instead of Distribution')
+
+groupHOG = parser.add_argument_group('HOG arguments:')
+groupHOG.add_argument('-HOG', action='store_true', help='Use option to activate HOG')
+groupHOG.add_argument('--HOG_CellD', metavar='INT', action='store', help='CellDimension for local histograms', type=int, default=5)
+groupHOG.add_argument('--HOG_Orient', metavar='INT', action='store', help='Number of bins of local histograms', type=int, default=8)
+groupHOG.add_argument('--HOG_Cluster', metavar='INT', action='store', help='Number of k-means cluster', type=int, default=12)
+groupHOG.add_argument('--HOG_Iter', metavar='INT', action='store', help='Max. number of iterations for clustering', type=int, default=100)
+groupHOG.add_argument('--HOG_cores', metavar='INT', action='store', help='Number of cores for HOG', type=int, default=1)
+
+### Read args
args = parser.parse_args()
-path = args.path
-NB_CORES = args.cores
-nameDB = args.name
-### Helper
+nameDB = args.name
+path = args.path
-# Function to transform the boolean deciscion of norm into a string
+### Helper-Function to transform the boolean deciscion of norm into a string
def boolNormToStr(norm):
if(norm):
return "MinMax"
@@ -92,7 +90,7 @@ if(args.SURF):
if(args.HOG):
features = features + "HOG"
-print "Infos:\t NameDB=" + nameDB + ", Path=" + path + ", Cores=" + str(NB_CORES) + ", Features=" + features
+print "Infos:\t NameDB=" + nameDB + ", Path=" + path + ", Features=" + features
################################ Read Images from Database
# Determine the Database to extract features
@@ -119,16 +117,12 @@ if(args.RGB):
print "RGB:\t Start"
t_rgb_start = time.time()
-
- numberOfBins = args.RGB_Hist
- maxColorIntensity = args.RGB_CI
- boolNormMinMax = args.RGB_NMinMax
-
+
# Infos
- print "RGB:\t NumberOfBins=" + str(numberOfBins) + ", MaxColorIntensity=" + str(maxColorIntensity) + ", Norm=" + boolNormToStr(boolNormMinMax)
+ print "RGB:\t NumberOfBins=" + str(args.RGB_Bins) + ", MaxColorIntensity=" + str(args.RGB_CI) + ", Norm=" + boolNormToStr(args.RGB_NMinMax)
# Extract Feature from DB
- rgb_feat_desc,rgb_f_extr_res = FeatExtraction.calcRGBColorHisto(nameDB, dfImages, numberOfBins, maxColorIntensity, boolNormMinMax)
+ rgb_feat_desc,rgb_f_extr_res = FeatExtraction.calcRGBColorHisto(nameDB, dfImages, args.RGB_Bins, args.RGB_CI, args.RGB_NMinMax)
t_rgb = time.time() - t_rgb_start
print "RGB:\t Done in: " + str(t_rgb) + "[s]"
@@ -139,17 +133,16 @@ if(args.HSV):
print "HSV:\t Start"
t_hsv_start = time.time()
- h_bins = args.HSV_H
- s_bins = args.HSV_S
- v_bins = args.HSV_V
+ h_bins = args.HSV_H_Bins
+ s_bins = args.HSV_S_Bins
+ v_bins = args.HSV_V_Bins
histSize = [h_bins, s_bins, v_bins]
- boolNormMinMax = args.HSV_NMinMax
# Infos
- print "HSV:\t HSVBins=[" + str(h_bins) + "," + str(s_bins) + "," + str(v_bins) + "], Norm=" + boolNormToStr(boolNormMinMax)
+ print "HSV:\t HSVBins=[" + str(h_bins) + "," + str(s_bins) + "," + str(v_bins) + "], Norm=" + boolNormToStr(args.HSV_NMinMax)
# Extract Feature from DB
- hsv_feat_desc,hsv_f_extr_res = FeatExtraction.calcHSVColorHisto(nameDB, dfImages, histSize, boolNormMinMax)
+ hsv_feat_desc,hsv_f_extr_res = FeatExtraction.calcHSVColorHisto(nameDB, dfImages, histSize, args.HSV_NMinMax)
t_hsv = time.time() - t_hsv_start
print "HSV:\t Done in: " + str(t_hsv) + "[s]"
@@ -161,13 +154,11 @@ if(args.SIFT):
t_sift_start = time.time()
boolSIFT = True
- cluster = args.SIFT_Cluster
- boolNormMinMax = args.SIFT_NMinMax
- print "SIFT:\t Cluster=" + str(cluster) + ", Norm=" + boolNormToStr(boolNormMinMax)
+ print "SIFT:\t Cluster=" + str(args.SIFT_Cluster) + ", Norm=" + boolNormToStr(args.SIFT_NMinMax)
sift_descriptors,sift_des_list = FeatExtraction.calcSURFSIFTDescriptors(dfImages, boolSIFT)
- sift_feat_desc,sift_f_extr_res = FeatExtraction.calcSURFSIFTHisto(nameDB, dfImages, cluster, boolNormMinMax, sift_descriptors, sift_des_list, boolSIFT)
+ sift_feat_desc,sift_f_extr_res = FeatExtraction.calcSURFSIFTHisto(nameDB, dfImages, args.SIFT_Cluster, args.SIFT_NMinMax, sift_descriptors, sift_des_list, boolSIFT)
t_sift = time.time() - t_sift_start
print "SIFT:\t Done in: " + str(t_sift) + "[s]"
@@ -178,14 +169,12 @@ if(args.SURF):
t_surf_start = time.time()
boolSIFT = False
- cluster = args.SURF_Cluster
- boolNormMinMax = args.SURF_NMinMax
- print "SURF:\t Cluster=" + str(cluster) + ", Norm=" + boolNormToStr(boolNormMinMax)
+ print "SURF:\t Cluster=" + str(args.SURF_Cluster) + ", Norm=" + boolNormToStr(args.SURF_NMinMax)
# Extract Feature from DB
surf_descriptors,surf_des_list = FeatExtraction.calcSURFSIFTDescriptors(dfImages, boolSIFT)
- surf_feat_desc,surf_f_extr_res = FeatExtraction.calcSURFSIFTHisto(nameDB, dfImages, cluster, boolNormMinMax, surf_descriptors, surf_des_list, boolSIFT)
+ surf_feat_desc,surf_f_extr_res = FeatExtraction.calcSURFSIFTHisto(nameDB, dfImages, args.SURF_Cluster, args.SURF_NMinMax, surf_descriptors, surf_des_list, boolSIFT)
t_surf = time.time() - t_surf_start
print "SURF:\t Done in: " + str(t_surf) + "[s]"
@@ -198,6 +187,7 @@ if(args.HOG):
NB_ORIENTATIONS = args.HOG_Orient
NB_CLUSTERS = args.HOG_Cluster
MAXITER = args.HOG_Iter
+ NB_CORES = args.HOG_cores
print "HOG:\t CellDim=" + str(CELL_DIMENSION) + ", NbOrientations=" + str(NB_ORIENTATIONS) +", Cluster=" + str(NB_CLUSTERS) + ", MaxIter=" + str(MAXITER)
diff --git a/Code/FeatExtraction/ExecFeatParaOpt.py b/Code/FeatExtraction/ExecFeatParaOpt.py
index 2211c07ae8b9e458c3b41fe48eaa36375ee3a0ee..5d5ed749629a53841c6c2a50c9d0178e6aa8ba3a 100644
--- a/Code/FeatExtraction/ExecFeatParaOpt.py
+++ b/Code/FeatExtraction/ExecFeatParaOpt.py
@@ -3,34 +3,94 @@
""" Script to perform feature parameter optimisation """
# Import built-in modules
-import cv2 # for OpenCV
-import cv # for OpenCV
import datetime # for TimeStamp in CSVFile
-from scipy.cluster.vq import * # for Clustering http://docs.scipy.org/doc/scipy/reference/cluster.vq.html
-import numpy as np # for arrays
-from argparse import ArgumentParser # for acommand line arguments
+import argparse # for acommand line arguments
-# Import sci-kit learn
-from sklearn.ensemble import RandomForestClassifier
+# Import 3rd party modules
+import numpy as np # for arrays
# Import own modules
-import DBCrawl # Functions to read Images from Database
-import FeatParaOpt # Functions to perform parameter optimisation
-import ExportResults # Functions to render results
+import DBCrawl # Functions to read Images from Database
+import FeatParaOpt # Functions to perform parameter optimisation
+import ExportResults # Functions to render results
# Author-Info
__author__ = "Nikolas Huelsmann"
-__status__ = "Development" #Production, Development, Prototype
+__status__ = "Development" #Production, Development, Prototype
__date__ = 2016-01-23
+### Argument Parser
+
+parser = argparse.ArgumentParser(
+description='This methods permits to perform an optimisation of the parameter of one feature. Therefore you have so specify which feature to use (e.g. --feature RGB) and which of his parameters (the parameters depend on the feature chosen, e.g. for RGB: --parameter Bins). The method will calculate the results in your given range and export the results to a CSV-File.',
+formatter_class=argparse.ArgumentDefaultsHelpFormatter)
+
+groupStandard = parser.add_argument_group('necessary arguments:')
+groupStandard.add_argument('--name', metavar='STRING', action='store', help='Select a name of DB, e.g. Caltech (default: %(default)s)', default='DB')
+groupStandard.add_argument('--path', metavar='STRING', action='store', help='Path to the database (default: %(default)s)', default='D:\\CaltechMini')
+
+groupOpt = parser.add_argument_group('Optimisation arguments:')
+groupOpt.add_argument('--feature', choices=['RGB', 'HSV', 'SURF', 'SIFT', 'HOG'], help='Set feature from list (RGB, HSV, ..)', default='RGB')
+groupOpt.add_argument('--param', choices=['RGB_Bins', 'RGB_MaxCI', 'HSV_H_Bins', 'HSV_S_Bins', 'HSV_V_Bins', 'SIFT_Cluster', 'SURF_Cluster', 'HOG_Cluster'], help='Parameter to optimise (remember depends on feature)', default='RGB_Bins')
+groupOpt.add_argument('--valueStart', metavar='INT', action='store', help='Start-Value for optimisation range', type=int)
+groupOpt.add_argument('--valueEnd', metavar='INT', action='store', help='End-Value for optimisation range', type=int)
+groupOpt.add_argument('--nCalcs', metavar='INT', action='store', help='Number of calculations between Start and End-Value', type=int)
+
+groupRGB = parser.add_argument_group('RGB arguments:')
+groupRGB.add_argument('--RGB_Bins', metavar='INT', action='store', help='Number of bins for histogram', type=int, default=16)
+groupRGB.add_argument('--RGB_CI', metavar='INT', action='store', help='Max Color Intensity [0 to VALUE]', type=int, default=256)
+groupRGB.add_argument('-RGB_NMinMax', action='store_true', help='Use option to actvate MinMax Norm instead of Distribution')
+
+groupHSV = parser.add_argument_group('HSV arguments:')
+groupHSV.add_argument('--HSV_H_Bins', metavar='INT', action='store', help='Number of bins for Hue', type=int, default=16)
+groupHSV.add_argument('--HSV_S_Bins', metavar='INT', action='store', help='Number of bins for Saturation', type=int, default=4)
+groupHSV.add_argument('--HSV_V_Bins', metavar='INT', action='store', help='Number of bins for Value', type=int, default=4)
+groupHSV.add_argument('-HSV_NMinMax', action='store_true', help='Use option to actvate MinMax Norm instead of Distribution')
+
+groupSIFT = parser.add_argument_group('SIFT arguments:')
+groupSIFT.add_argument('--SIFT_Cluster', metavar='INT', action='store', help='Number of k-means cluster', type=int, default=50)
+groupSIFT.add_argument('-SIFT_NMinMax', action='store_true', help='Use option to actvate MinMax Norm instead of Distribution')
+
+groupSURF = parser.add_argument_group('SURF arguments:')
+groupSURF.add_argument('--SURF_Cluster', metavar='INT', action='store', help='Number of k-means cluster', type=int, default=50)
+groupSURF.add_argument('-SURF_NMinMax', action='store_true', help='Use option to actvate MinMax Norm instead of Distribution')
+
+groupHOG = parser.add_argument_group('HOG arguments:')
+groupHOG.add_argument('--HOG_CellD', metavar='INT', action='store', help='CellDimension for local histograms', type=int, default=5)
+groupHOG.add_argument('--HOG_Orient', metavar='INT', action='store', help='Number of bins of local histograms', type=int, default=8)
+groupHOG.add_argument('--HOG_Cluster', metavar='INT', action='store', help='Number of k-means cluster', type=int, default=12)
+groupHOG.add_argument('--HOG_Iter', metavar='INT', action='store', help='Max. number of iterations for clustering', type=int, default=100)
+groupHOG.add_argument('--HOG_cores', metavar='INT', action='store', help='Number of cores for HOG', type=int, default=1)
+
+groupClass = parser.add_argument_group('Classification arguments:')
+groupClass.add_argument('--CL_split', metavar='DOUBLE', action='store', help='Determine the the train/test split', type=double, default=0.7)
+groupClass.add_argument('--CL_RF_trees', metavar='STRING', action='store', help='GridSearch: Determine the trees', default='[50, 100, 150, 200]')
+groupClass.add_argument('--CL_RF_CV', metavar='INT', action='store', help='Number of k-folds for CV', type=int, default=3)
+groupClass.add_argument('--CL_RF_Cores', metavar='INT', action='store', help='Number of cores', type=int, default=1)
+
+
+
+### Read args - transform in Arrays for function calls
+args = parser.parse_args()
+path = args.path
+nameDB = args.name
+
+para_opt = [args.feature, args.param, args.valueStart, args.valueEnd, args.nCalcs]
+para_RGB = [args.RGB_Bins, args.RGB_CI, args.RGB_NMinMax]
+para_HSV = [args.HSV_H_Bins, args.HSV_S_Bins, args.HSV_V_Bins, args.HSV_NMinMax]
+para_SIFT = [args.SIFT_Cluster, args.SIFT_NMinMax]
+para_SURF = [args.SURF_Cluster, args.SURF_NMinMax]
+para_HOG = [args.HOG_CellD, args.HOG_Orient, args.HOG_Cluster, args.HOG_Iter, args.HOG_cores]
+para_Cl = [args.CL_split, args.CL_RF_trees, args.CL_RF_CV, args.CL_RF_Cores]
+
+
### Main Programm
################################ Read Images from Database
# Determine the Database to extract features
print "### Start of Main Programm for Feature Parameter Optimisation ###"
-path ="D:\\Caltech"
-nameDB = "CT"
+
print "### Start:\t Exportation of images from DB ###"
@@ -45,32 +105,12 @@ print "### Done:\t Exportation of Images from DB ###"
################################ Parameter Optimisation
# Setup
-#feature = "RGB"
-#parameter = "Bins"
-#valueStart = int(8)
-#valueEnd = int(64)
-#nCalculations = int(8)
-#boolCV = True
-
-#print '### Optimisation - Feature:' + str(feature) + " Parameter:" + str(parameter) + " from:" + str(valueStart) + " to:" + str(valueEnd) + " in #calc:" + str(nCalculations) + " withCV:" + str(boolCV) + " ###"
-
-#print "### Start: Feautre Optimisation ###"
-#df_feat_res = FeatParaOpt.perfFeatMonoV(nameDB, dfImages,feature, parameter, valueStart, valueEnd, nCalculations, boolCV)
-#print "### Done: Feautre Optimisation ###"
-
-# Setup SURF
-feature = "SURF"
-parameter = "Cluster"
-valueStart = 50
-valueEnd = 200
-nCalculations = 4
-boolCV = True
-
-print '### Optimisation - Feature:' + str(feature) + " Parameter:" + str(parameter) + " from:" + str(valueStart) + " to:" + str(valueEnd) + " in #calc:" + str(nCalculations) + " withCV:" + str(boolCV) + " ###"
-
-print "### Start:\t Feautre Optimisation ###"
-df_feat_res = FeatParaOpt.perfFeatMonoV(nameDB, dfImages,feature, parameter, valueStart, valueEnd, nCalculations, boolCV)
-print "### Done:\t Feautre Optimisation ###"
+print '### Optimisation - Feature:' + str(args.feature) + " Parameter:" + str(args.param) + " from:" + str(args.valueStart) + " to:" + str(args.valueEnd) + " in #calc:" + str(args.nCalcs) + " ###"
+
+print "### Start: Feautre Optimisation ###"
+df_feat_res = FeatParaOpt.perfFeatMonoV(nameDB, dfImages, para_opt, para_RGB, para_HSV, para_SIFT, para_SURF, para_HOG, para_Cl)
+print "### Done: Feautre Optimisation ###"
+
################################ Render results
print "### Start:\t Exporting to CSV ###"
diff --git a/Code/FeatExtraction/ExportResults.py b/Code/FeatExtraction/ExportResults.py
index f1a1209680469b4af8a432099d543bf1982b1dd8..99f797be1044b5b226d88eace67dfdb9da13cb1a 100644
--- a/Code/FeatExtraction/ExportResults.py
+++ b/Code/FeatExtraction/ExportResults.py
@@ -4,14 +4,14 @@
# Import built-in modules
import os # for iteration throug directories
+import string # to generate a range of letters
+
+# Import 3rd party modules
import pandas as pd # for Series and DataFrames
import numpy as np # for Numpy Arrays
import matplotlib.pyplot as plt # for Plots
from scipy.interpolate import interp1d # to Interpolate Data
-import string # to generate a range of letters
-from matplotlib.offsetbox import AnchoredOffsetbox, TextArea, DrawingArea, HPacker # to generate the Annotations in plot
-
-# Import 3rd party modules
+from matplotlib.offsetbox import AnchoredOffsetbox, TextArea, HPacker # to generate the Annotations in plot
# Import own modules
diff --git a/Code/FeatExtraction/FeatExtraction.py b/Code/FeatExtraction/FeatExtraction.py
index a716112715fbc1cdf2e4e6cfaac9208ded14d877..5ab462e6b2748151377eadda31ac31076d4addce 100644
--- a/Code/FeatExtraction/FeatExtraction.py
+++ b/Code/FeatExtraction/FeatExtraction.py
@@ -3,13 +3,11 @@
""" Code to Extract all Features from Database """
# Import built-in modules
-import numpy as np # for numpy arrays
-import pandas as pd # for panda DataFrame
-import datetime # for time calculations
-
-# Import OpenCV
+
+
+# Import 3rd party modules
+import numpy as np # for numpy arrays
import cv2 # for OpenCV
-import cv # for SIFT/SURF
from scipy.cluster.vq import * # for Clustering http://docs.scipy.org/doc/scipy/reference/cluster.vq.html\n",
# Import own modules
@@ -49,7 +47,6 @@ def calcRGBColorHisto(nameDB, dfImages, numberOfBins, maxColorIntensity, boolNor
param = "Bins_" + str(int(numberOfBins)) + "-" + "MaxCI_" + str(maxColorIntensity) + "-" + "Norm_" + norm
description = nameDB + "-RGB-" + param
-
## algo
for images in npImages:
diff --git a/Code/FeatExtraction/FeatParaOpt.py b/Code/FeatExtraction/FeatParaOpt.py
index a708b77969661ca9cb1f1c48279e101c7b8ae953..e846cb95fa74931d5ef71cca6670d542889b232f 100644
--- a/Code/FeatExtraction/FeatParaOpt.py
+++ b/Code/FeatExtraction/FeatParaOpt.py
@@ -3,11 +3,11 @@
""" Function to optimise feature parameters """
# Import built-in modules
-import time # for time calculations
-import numpy as np# for numpy arrays
-import pandas as pd # for Series and DataFrames
+import time # for time calculations
# Import 3rd party modules
+import numpy as np # for numpy arrays
+import pandas as pd # for Series and DataFrames
# Import own modules
import FeatExtraction # Functions for Feature Extractions#
@@ -16,22 +16,28 @@ import ClassifMonoView # Functions for classification
# Author-Info
__author__ = "Nikolas Huelsmann"
__status__ = "Development" #Production, Development, Prototype
-__date__ = 2016-01-23
+__date__ = 2016-03-14
-
-# dfImages: Database with all images
-# feature: which feature? e.g. ColorHistogram
-# paramter: which parameter should be tested? e.g. bins of histogram
-# valueStart: Value for paramter to start with
-# valueEnd: Value for paramter to end with
-# nCalculations: How many calculations between valueStart and valueEnd? e.g. vS=0,VE=9,nCalc=10 -> test:0,1,2,3,4,5,6,7,8,9
-# boolCV: Boolian if CrossValidation should be used
-def perfFeatMonoV(nameDB, dfImages,feature, parameter, valueStart, valueEnd, nCalculations, boolCV):
+# dfImages: Database with all images
+# feature: which feature? e.g. ColorHistogram
+# para_opt: optimisation parameters
+# para_RGB: RGB parameters
+# para_HSV: HSV paramters
+# para_SIFT: SIFT parameters
+# para_SURF: SURF parameters
+# para_HOG: HOG paramters
+# para_Cl: Classification parameters
+def perfFeatMonoV(nameDB, dfImages, para_opt, para_RGB, para_HSV, para_SIFT, para_SURF, para_HOG, para_Cl):
# TIME for total calculation
t_tot_start = time.time()
- # Value check - are the given values possible: e.g. bins valueStart = -1 -> error
+ # Values from Array into variables - easier to read the code
+ feature = para_opt[0]
+ parameter = para_opt[1]
+ valueStart = para_opt[2]
+ valueEnd = para_opt[3]
+ nCalculations = para_opt[4]
# Calculate Stepwidth
if(nCalculations>1):
@@ -39,6 +45,7 @@ def perfFeatMonoV(nameDB, dfImages,feature, parameter, valueStart, valueEnd, nCa
valueArray = np.around(np.array(range(0,nCalculations))*step) + valueStart
else:
valueArray = [valueStart]
+
# FeatExtraction Results DataFrame
df_feat_res = pd.DataFrame()
@@ -58,49 +65,53 @@ def perfFeatMonoV(nameDB, dfImages,feature, parameter, valueStart, valueEnd, nCa
# Call extraction function with parameters -> returns feature
if(feature=="RGB"):
# Basic Setup
- numberOfBins = 16
- maxColorIntensity = 256
- boolNormMinMax = False
+ numberOfBins = para_RGB[0]
+ maxColorIntensity = para_RGB[1]
+ boolNormMinMax = para_RGB[2]
+
# ParamaterTest
- if(parameter=="Bins"):
+ if(parameter=="RGB_Bins"):
numberOfBins = valuePara
- elif(parameter=="MaxCI"):
+ elif(parameter=="RGB_MaxCI"):
maxColorIntensity = valuePara
- elif(parameter=="Norm"):
- boolNormMinMax = valuePara
# Extract Feature from DB
feat_desc,f_extr_res = FeatExtraction.calcRGBColorHisto(nameDB, dfImages, numberOfBins, maxColorIntensity, boolNormMinMax)
elif(feature=="HSV"):
# Basic Setup
- h_bins = 8
- s_bins = 3
- v_bins = 3
- histSize = [h_bins, s_bins, v_bins]
- boolNormMinMax = False
+ h_bins = para_HSV[0]
+ s_bins = para_HSV[1]
+ v_bins = para_HSV[2]
+
+ boolNormMinMax = para_HSV[3]
+
+ HSV_H_Bins
# ParamaterTest
- if(parameter=="Bins"):
- histSize = valuePara
- elif(parameter=="Norm"):
- boolNormMinMax = valuePara
-
+ if(parameter=="HSV_H_Bins"):
+ h_bins = valuePara
+ elif(parameter=="HSV_S_Bins"):
+ s_bins = valuePara
+ elif(parameter=="HSV_V_Bins"):
+ v_bins = valuePara
+
+ histSize = [h_bins, s_bins, v_bins]
+
# Extract Feature from DB
feat_desc,f_extr_res = FeatExtraction.calcHSVColorHisto(nameDB, dfImages, histSize, boolNormMinMax)
elif(feature=="SURF"):
# Basic Setup
- cluster = 50
- boolNormMinMax = False
+ cluster = para_SIFT[0]
+ boolNormMinMax = para_SIFT[1]
boolSIFT = False
+
# ParamaterTest
- if(parameter=="Cluster"):
+ if(parameter=="SIFT_Cluster"):
cluster = valuePara
- elif(parameter=="Norm"):
- boolNormMinMax = valuePara
if descriptors is None:
descriptors,des_list = FeatExtraction.calcSURFSIFTDescriptors(dfImages, boolSIFT)
@@ -110,21 +121,34 @@ def perfFeatMonoV(nameDB, dfImages,feature, parameter, valueStart, valueEnd, nCa
elif(feature=="SIFT"):
# Basic Setup
- cluster = 50
- boolNormMinMax = False
+ cluster = para_SURF[0]
+ boolNormMinMax = para_SURF[1]
boolSIFT = True
+
# ParamaterTest
- if(parameter=="Cluster"):
+ if(parameter=="SURF_Cluster"):
cluster = valuePara
- elif(parameter=="Norm"):
- boolNormMinMax = valuePara
if descriptors is None:
descriptors,des_list = FeatExtraction.calcSURFSIFTDescriptors(dfImages, boolSIFT)
# Extract Feature from DB
feat_desc,f_extr_res = FeatExtraction.calcSURFSIFTHisto(nameDB, dfImages, cluster, boolNormMinMax, descriptors, des_list, boolSIFT)
+ elif(feature=="HOG"):
+ CELL_DIMENSION = para_HOG[0]
+ NB_ORIENTATIONS = para_HOG[1]
+ NB_CLUSTERS = para_HOG[2]
+ MAXITER = para_HOG[3]
+ NB_CORES = para_HOG[4]
+
+ # ParamaterTest
+ if(parameter=="HOG_Cluster"):
+ NB_CLUSTERS = valuePara
+
+ # Extract Feature from DB
+ feat_desc,f_extr_res = FeatExtraction.calcHOGParallel(nameDB, npImages, CELL_DIMENSION, NB_ORIENTATIONS, NB_CLUSTERS, MAXITER, NB_CORES):
+
else:
print "ERROR: Selected Feature does not exist"
print "Feature: " + str(feature)
@@ -137,26 +161,23 @@ def perfFeatMonoV(nameDB, dfImages,feature, parameter, valueStart, valueEnd, nCa
# TIME for CLASSIFICATION
t_classif_start = time.time()
+ # Values from Array into variables - easier to read the code
+ split = para_Cl[0]
+ num_estimators = para_Cl[1]
+ cv_folds = para_Cl[2]
+ clas_cores = para_Cl[3]
+
# Calculate Train/Test data
- #Basic Setup
- split = 0.7
X_train, X_test, y_train, y_test = ClassifMonoView.calcTrainTest(f_extr_res, dfImages.classLabel, split)
-
# Own Function for split: ClassifMonoView.calcTrainTestOwn
# Begin Classification RandomForest
# call function: return fitted model
print "### Start:\t Classification Nr:" + str(i) + " from:" + str(max(arr_Calc)) + " ###"
-
- # Basic Setup
- num_estimators = [50, 101, 150, 200]
-
- if(boolCV==True):
- cl_desc, cl_res = ClassifMonoView.calcClassifRandomForestCV(X_train, y_train, num_estimators)
- else:
- cl_desc, cl_res = ClassifMonoView.calcClassifRandomForest(X_train, X_test, y_test, y_train, num_estimators)
-
+
+ cl_desc, cl_res = ClassifMonoView.calcClassifRandomForestCV(X_train, y_train, num_estimators, cv_folds, clas_cores)
+
print "### Done:\t Classification Nr:" + str(i) + " from:" + str(max(arr_Calc)) + " ###"
# TIME for CLASSIFICATION END
diff --git a/Code/FeatExtraction/Versions.py b/Code/FeatExtraction/Versions.py
new file mode 100644
index 0000000000000000000000000000000000000000..90392028c4f557fd124f686406b17de3a252aedf
--- /dev/null
+++ b/Code/FeatExtraction/Versions.py
@@ -0,0 +1,38 @@
+#!/usr/bin/env python
+
+""" Script to render versions of modules used """
+
+# Import built-in modules
+
+# Import 3rd party modules
+
+# Import own modules
+
+
+# Author-Info
+__author__ = "Nikolas Huelsmann"
+__status__ = "Development" #Production, Development, Prototype
+__date__ = 2016-03-14
+
+import sys
+print("Python-V.: " + sys.version)
+
+import cv2
+print("OpenCV2-V.: " + cv2.__version__)
+
+import pandas as pd
+print("Pandas-V.: " + pd.__version__)
+
+import numpy
+print("Numpy-V.: " + numpy.version.version)
+
+import scipy
+print("Scipy-V.: " + scipy.__version__)
+
+import matplotlib
+print("Matplotlib-V.: " + matplotlib.__version__)
+
+import sklearn
+print("Sklearn-V.: " + sklearn.__version__)
+
+