filter_name = raw_input('Filter name (return for none): ')
if filter_name == '':
filter_name = None
group = raw_input('Group name (return for none): ')
if group == '':
group = 'Image'
results_table = raw_input('Results table name (return for none): ')
logging.info('Loading properties file...')
p = Properties.getInstance()
p.LoadFile(props_file)
logging.info('Loading training set...')
ts = TrainingSet(p)
ts.Load(ts_file)
score(p,
ts,
nRules,
filter_name,
group,
show_results=True,
results_table=results_table,
overwrite=False)
app.MainLoop()
#
# Kill the Java VM
db = DBConnect.getInstance()
dm = DataModel.getInstance()
# props = '/Volumes/imaging_analysis/2007_10_19_Gilliland_LeukemiaScreens/Screen3_1Apr09_run3/2007_10_19_Gilliland_LeukemiaScreens_Validation_v2_AllBatches_DuplicatesFiltered_FullBarcode_testSinglePlate.properties'
# ts = '/Volumes/imaging_analysis/2007_10_19_Gilliland_LeukemiaScreens/Screen3_1Apr09_run3/trainingvalidation3b.txt'
props = '../Properties/nirht_area_test.properties'
ts = '/Users/afraser/Desktop/MyTrainingSet3.txt'
nRules = 5
filter = 'MAPs'
# props = '/Users/afraser/Desktop/2007_10_19_Gilliland_LeukemiaScreens_Validation_v2_AllBatches_DuplicatesFiltered_FullBarcode.properties'
# ts = '/Users/afraser/Desktop/trainingvalidation3d.txt'
# nRules = 50
# filter = 'afraser_test'
p.LoadFile(props)
trainingSet = TrainingSet(p)
trainingSet.Load(ts)
output = StringIO()
print('Training classifier with ' + str(nRules) + ' rules...')
weaklearners = fastgentleboostingmulticlass.train(trainingSet.colnames,
nRules,
trainingSet.label_matrix,
trainingSet.values,
output)
table = PerImageCounts(weaklearners, filter_name=filter)
table.sort()
labels = ['table', 'image'] + list(trainingSet.labels) + list(
trainingSet.labels)
print(labels)
for row in table:
import glob, os
from PIL import Image
from trainingset import TrainingSet
# Declare directories
png_directory = "/Users/jessetvogel/Downloads/mnist_png"
nnts_directory = "/Users/jessetvogel/Desktop/data"
# Scan directories
for subdirectory in ["training", "testing"]:
print "Scanning " + subdirectory + " directory ..."
training_set = TrainingSet(1, 1, 0.0, 1.0, 0.0, 1.0, 28 * 28, 10)
for i in range(10):
print "Scanning for digit " + str(i)
n = 0
directory = png_directory + "/" + subdirectory + "/" + str(i)
for file in os.listdir(directory):
if file.endswith(".png"):
img = Image.open(directory + "/" + file)
output = [0.0] * 10
output[i] = 1.0
training_set.add_sample([x / 255.0 for x in list(img.getdata())], output)
img.close()
n += 1
if n == 25:
break
print "Read " + str(n) + " images"
print "Saving " + subdirectory + " ..."
training_set.save(nnts_directory + "/" + subdirectory + ".nnts")
print ""
##
#group = raw_input('Group name (return for none): ')
#if group=='':
#group = 'Image'
group = 'None'
##
#results_table = raw_input('Results table name (return for none): ')
results_table = ''
logging.info('Loading properties file...')
p = Properties.getInstance()
p.LoadFile(props_file)
logging.info('Loading initial training set...')
ts = TrainingSet(p)
ts.Load(ts_file)
logging.info('Loading ground truth training set...')
gt = TrainingSet(p)
gt.Load(gt_file)
score_objects(p,
ts,
gt,
nRules,
filter_name,
group,
show_results=True,
results_table=results_table,
overwrite=False)
from trainingset import TrainingSet path = "tmp_file.nnts" # Create training set tset = TrainingSet(1, 1, 0.0, 1.0, 0.0, 1.0, 2, 1) tset.add_sample([0.0, 0.0], [0.0]) tset.add_sample([0.0, 1.0], [1.0]) tset.add_sample([1.0, 0.0], [1.0]) tset.add_sample([1.0, 1.0], [0.0]) # Write training set tset.save(path) # # Clean up # import os # os.remove(path)