def classify_image(cc, img_idx, C=1.0, gamma=0.0, cls=None):
"""
Input: ClassifierConfig, img_idx
Output: Score for given class; 20list of scores if cls=None
"""
feat_vect = get_feature_vector(cc, img_idx)
if cls == None:
score = {}
for cls_idx, cls in enumerate(cc.d.classes):
print '\tclassify %s for %s on %d'%(cc.d.images[img_idx].name, cls, comm_rank)
path = config.get_classifier_svm_name(cls, C, gamma)
if os.path.exists(path):
clf = load_svm(path)
score[cls_idx] = svm_predict(feat_vect, clf)
else:
clf = load_svm(config.get_classifier_svm_name(cls))
score = svm_predict(feat_vect, clf)
return score
def classify_image(cc, img_idx, C=1.0, gamma=0.0, cls=None):
"""
Input: ClassifierConfig, img_idx
Output: Score for given class; 20list of scores if cls=None
"""
feat_vect = get_feature_vector(cc, img_idx)
if cls == None:
score = {}
for cls_idx, cls in enumerate(cc.d.classes):
print '\tclassify %s for %s on %d' % (cc.d.images[img_idx].name,
cls, comm_rank)
path = config.get_classifier_svm_name(cls, C, gamma)
if os.path.exists(path):
clf = load_svm(path)
score[cls_idx] = svm_predict(feat_vect, clf)
else:
clf = load_svm(config.get_classifier_svm_name(cls))
score = svm_predict(feat_vect, clf)
return score
def classify_images(cc, images, C, gamma):
res = np.zeros((images.shape[0], len(cc.d.classes)))
for cls_idx, cls in enumerate(cc.d.classes):
filename = config.get_classifier_svm_name(cls, C, gamma, cc.d.current_fold)
if not os.path.exists(filename):
continue
clf = load_svm(filename, probability=False)
for idx2, img_idx in enumerate(images):
x = get_feature_vector(cc, img_idx)
pred = svm_predict(x, clf)
if pred.size > 0:
res[idx2, cls_idx] = 1
return res*2-1
def classify_images(cc, images, C, gamma):
res = np.zeros((images.shape[0], len(cc.d.classes)))
for cls_idx, cls in enumerate(cc.d.classes):
filename = config.get_classifier_svm_name(cls, C, gamma,
cc.d.current_fold)
if not os.path.exists(filename):
continue
clf = load_svm(filename, probability=False)
for idx2, img_idx in enumerate(images):
x = get_feature_vector(cc, img_idx)
pred = svm_predict(x, clf)
if pred.size > 0:
res[idx2, cls_idx] = 1
return res * 2 - 1