from settings import argparse_settings
sett = argparse_settings("Train detector")
dsettings = sett['detector']
#import argparse
#parser = argparse.ArgumentParser(description='Train mixture model on edge data')
#parser.add_argument('patches', metavar='<patches file>', type=argparse.FileType('rb'), help='Filename of patches file')
#parser.add_argument('model', metavar='<output model file>', type=argparse.FileType('wb'), help='Filename of the output models file')
#parser.add_argument('mixtures', metavar='<number mixtures>', type=int, help='Number of mixture components')
#parser.add_argument('--use-voc', action='store_true', help="Use VOC data to train model")
import gv
import glob
import os
import os.path
import amitgroup as ag
ag.set_verbose(True)
#descriptor = gv.load_descriptor(gv.BinaryDetector.DESCRIPTOR, sett)
descriptor = gv.load_descriptor(sett)
detector = gv.BernoulliDetector(dsettings['num_mixtures'], descriptor, dsettings)
if dsettings['use_voc']:
files = gv.voc.load_object_images_of_size(sett['voc'], 'bicycle', dsettings['image_size'], dataset='train')
else:
base_path = ''
from __future__ import division
from settings import argparse_settings
sett = argparse_settings("Check prevalence")
esettings = sett['edges']
dsettings = sett['parts']
import numpy as np
import matplotlib.pylab as plt
import amitgroup as ag
import glob
import os
import sys
import gv
descriptor = gv.BinaryDescriptor.getclass('edges').load_from_dict(esettings)
base_path = dsettings.get('base_path')
path = dsettings['image_dir']
if base_path is not None:
path = os.path.join(os.environ[base_path], path)
files = glob.glob(path)[:1]
counts = np.zeros(8)
tots = 0
def process(f):
#print "Processing file {0}".format(f)
im = gv.img.load_image(f)
edges = descriptor.extract_features(im, {'radius': 0})
ag.info('{0} Farming {1}'.format(i, fileobj.img_id))
img = gv.img.load_image(fileobj.path)
grayscale_img = gv.img.asgray(img)
bbobjs = detector.detect_coarse(grayscale_img, fileobj=fileobj, mixcomps=[mixcomp], use_padding=False, use_scale_prior=False, cascade=True, discard_weak=True, more_detections=True, farming=True, save_samples=True)
for bbobj in bbobjs:
bbobj.img_id = fileobj.img_id
if bbobj.confidence > threshold:
topsy.append(bbobj)
return topsy
if gv.parallel.main(__name__):
from settings import argparse_settings
settings = argparse_settings("Train real-valued detector")
dsettings = settings['detector']
#descriptor = gv.load_descriptor(gv.RealDetector.DESCRIPTOR, sett)
descriptor = gv.load_descriptor(settings)
detector = gv.RealDetector(descriptor, dsettings)
files = get_training_files(detector)
neg_files = sorted(glob.glob(os.path.expandvars(dsettings['neg_dir'])))[:dsettings.get('neg_limit')]
#pos_images = []
image_size = detector.settings['image_size']
print('Edge type:', descriptor.settings.get('edge_type', '(none)'))
# Extract clusters (manual or through EM)
from settings import argparse_settings
sett = argparse_settings("Train detector")
dsettings = sett['detector']
#import argparse
#parser = argparse.ArgumentParser(description='Train mixture model on edge data')
#parser.add_argument('patches', metavar='<patches file>', type=argparse.FileType('rb'), help='Filename of patches file')
#parser.add_argument('model', metavar='<output model file>', type=argparse.FileType('wb'), help='Filename of the output models file')
#parser.add_argument('mixtures', metavar='<number mixtures>', type=int, help='Number of mixture components')
#parser.add_argument('--use-voc', action='store_true', help="Use VOC data to train model")
import gv
import glob
import os
import os.path
import amitgroup as ag
ag.set_verbose(True)
#descriptor = gv.load_descriptor(gv.BinaryDetector.DESCRIPTOR, sett)
descriptor = gv.load_descriptor(sett)
detector = gv.BernoulliDetector(dsettings['num_mixtures'], descriptor,
dsettings)
if dsettings['use_voc']:
files = gv.voc.load_object_images_of_size(sett['voc'],
'bicycle',
dsettings['image_size'],
dataset='train')
from settings import argparse_settings
sett = argparse_settings("Train parts")
psettings = sett[sett['detector']['descriptor']]
import parser
import gv
import amitgroup as ag
import os.path
import random
import glob
ag.set_verbose(True)
if gv.parallel.main(__name__):
path = os.path.expandvars(psettings['image_dir'])
files = glob.glob(path)
random.seed(0)
random.shuffle(files)
settings = dict(samples_per_image=psettings['samples_per_image'])
cls = gv.BinaryDescriptor.getclass(sett['detector']['descriptor'])
codebook = cls(psettings['part_size'], psettings['num_parts'], settings=psettings)
codebook.train_from_images(files[:psettings['num_images']])
print('num parts', codebook.num_parts)
codebook.save(psettings['file'])
from __future__ import division
from settings import argparse_settings
sett = argparse_settings("Check prevalence")
esettings = sett['edges']
dsettings = sett['parts']
import numpy as np
import matplotlib.pylab as plt
import amitgroup as ag
import glob
import os
import sys
import gv
descriptor = gv.BinaryDescriptor.getclass('edges').load_from_dict(esettings)
base_path = dsettings.get('base_path')
path = dsettings['image_dir']
if base_path is not None:
path = os.path.join(os.environ[base_path], path)
files = glob.glob(path)[:1]
counts = np.zeros(8)
tots = 0
def process(f):
#print "Processing file {0}".format(f)
im = gv.img.load_image(f)
edges = descriptor.extract_features(im, {'radius': 0})
