opencv/samples/python2/digits_adjust.py

163 lines
5.5 KiB
Python
Raw Normal View History

2012-06-28 19:33:11 +02:00
'''
Digit recognition adjustment.
Grid search is used to find the best parameters for SVM and KNearest classifiers.
SVM adjustment follows the guidelines given in
2012-06-28 19:33:11 +02:00
http://www.csie.ntu.edu.tw/~cjlin/papers/guide/guide.pdf
Threading or cloud computing (with http://www.picloud.com/)) may be used
2012-06-28 19:33:11 +02:00
to speedup the computation.
Usage:
digits_adjust.py [--model {svm|knearest}] [--cloud] [--env <PiCloud environment>]
2012-06-28 19:33:11 +02:00
--model {svm|knearest} - select the classifier (SVM is the default)
2012-07-02 13:04:43 +02:00
--cloud - use PiCloud computing platform
2012-06-28 19:33:11 +02:00
--env - cloud environment name
'''
# TODO cloud env setup tutorial
import numpy as np
import cv2
from multiprocessing.pool import ThreadPool
from digits import *
try:
2012-07-02 13:04:43 +02:00
import cloud
have_cloud = True
except ImportError:
have_cloud = False
2012-07-02 13:04:43 +02:00
2012-06-28 19:33:11 +02:00
def cross_validate(model_class, params, samples, labels, kfold = 3, pool = None):
n = len(samples)
folds = np.array_split(np.arange(n), kfold)
def f(i):
model = model_class(**params)
test_idx = folds[i]
train_idx = list(folds)
train_idx.pop(i)
train_idx = np.hstack(train_idx)
train_samples, train_labels = samples[train_idx], labels[train_idx]
test_samples, test_labels = samples[test_idx], labels[test_idx]
model.train(train_samples, train_labels)
resp = model.predict(test_samples)
score = (resp != test_labels).mean()
print ".",
return score
if pool is None:
scores = map(f, xrange(kfold))
else:
scores = pool.map(f, xrange(kfold))
return np.mean(scores)
2012-07-02 13:04:43 +02:00
class App(object):
def __init__(self, usecloud=False, cloud_env=''):
if usecloud and not have_cloud:
print 'warning: cloud module is not installed, running locally'
2012-06-28 19:33:11 +02:00
usecloud = False
2012-07-02 13:04:43 +02:00
self.usecloud = usecloud
self.cloud_env = cloud_env
if self.usecloud:
print 'uploading dataset to cloud...'
cloud.files.put(DIGITS_FN)
self.preprocess_job = cloud.call(self.preprocess, _env=self.cloud_env)
else:
self._samples, self._labels = self.preprocess()
2012-06-28 19:33:11 +02:00
2012-07-02 13:04:43 +02:00
def preprocess(self):
if self.usecloud:
cloud.files.get(DIGITS_FN)
digits, labels = load_digits(DIGITS_FN)
shuffle = np.random.permutation(len(digits))
digits, labels = digits[shuffle], labels[shuffle]
digits2 = map(deskew, digits)
2012-07-02 16:12:34 +02:00
samples = preprocess_hog(digits2)
2012-07-02 13:04:43 +02:00
return samples, labels
def get_dataset(self):
if self.usecloud:
return cloud.result(self.preprocess_job)
else:
return self._samples, self._labels
def run_jobs(self, f, jobs):
if self.usecloud:
jids = cloud.map(f, jobs, _env=self.cloud_env, _profile=True, _depends_on=self.preprocess_job)
ires = cloud.iresult(jids)
else:
pool = ThreadPool(processes=cv2.getNumberOfCPUs())
ires = pool.imap_unordered(f, jobs)
return ires
2012-07-02 13:04:43 +02:00
def adjust_SVM(self):
2012-07-02 16:12:34 +02:00
Cs = np.logspace(0, 10, 15, base=2)
gammas = np.logspace(-7, 4, 15, base=2)
2012-07-02 13:04:43 +02:00
scores = np.zeros((len(Cs), len(gammas)))
scores[:] = np.nan
print 'adjusting SVM (may take a long time) ...'
def f(job):
i, j = job
samples, labels = self.get_dataset()
params = dict(C = Cs[i], gamma=gammas[j])
score = cross_validate(SVM, params, samples, labels)
return i, j, score
2012-07-02 13:04:43 +02:00
ires = self.run_jobs(f, np.ndindex(*scores.shape))
for count, (i, j, score) in enumerate(ires):
scores[i, j] = score
print '%d / %d (best error: %.2f %%, last: %.2f %%)' % (count+1, scores.size, np.nanmin(scores)*100, score*100)
print scores
2012-07-02 16:12:34 +02:00
print 'writing score table to "svm_scores.npz"'
np.savez('svm_scores.npz', scores=scores, Cs=Cs, gammas=gammas)
2012-07-02 13:04:43 +02:00
i, j = np.unravel_index(scores.argmin(), scores.shape)
best_params = dict(C = Cs[i], gamma=gammas[j])
print 'best params:', best_params
print 'best error: %.2f %%' % (scores.min()*100)
return best_params
def adjust_KNearest(self):
print 'adjusting KNearest ...'
def f(k):
samples, labels = self.get_dataset()
err = cross_validate(KNearest, dict(k=k), samples, labels)
return k, err
best_err, best_k = np.inf, -1
for k, err in self.run_jobs(f, xrange(1, 9)):
if err < best_err:
best_err, best_k = err, k
print 'k = %d, error: %.2f %%' % (k, err*100)
best_params = dict(k=best_k)
print 'best params:', best_params, 'err: %.2f' % (best_err*100)
return best_params
2012-06-28 19:33:11 +02:00
if __name__ == '__main__':
import getopt
import sys
2012-06-28 19:33:11 +02:00
print __doc__
args, _ = getopt.getopt(sys.argv[1:], '', ['model=', 'cloud', 'env='])
args = dict(args)
args.setdefault('--model', 'svm')
args.setdefault('--env', '')
if args['--model'] not in ['svm', 'knearest']:
print 'unknown model "%s"' % args['--model']
sys.exit(1)
t = clock()
2012-07-02 13:04:43 +02:00
app = App(usecloud='--cloud' in args, cloud_env = args['--env'])
2012-06-28 19:33:11 +02:00
if args['--model'] == 'knearest':
2012-07-02 13:04:43 +02:00
app.adjust_KNearest()
2012-06-28 19:33:11 +02:00
else:
2012-07-02 13:04:43 +02:00
app.adjust_SVM()
2012-06-28 19:33:11 +02:00
print 'work time: %f s' % (clock() - t)