def test_load_save():
ndx1 = create_ndx()
file_h5 = output_dir + '/test.h5'
ndx1.save(file_h5)
ndx3 = TrialNdx.load(file_h5)
assert (ndx1 == ndx3)
file_txt = output_dir + '/test.txt'
ndx3.trial_mask[0, :] = True
ndx3.trial_mask[:, 0] = True
ndx3.save(file_txt)
ndx2 = TrialNdx.load(file_txt)
assert (ndx3 == ndx2)
def test_filter():
ndx1 = create_ndx()
ndx2 = TrialNdx(ndx1.model_set[:5], ndx1.seg_set[:10],
ndx1.trial_mask[:5, :10])
ndx3 = ndx1.filter(ndx2.model_set, ndx2.seg_set, keep=True)
assert (ndx2 == ndx3)
def test_split():
ndx1 = create_ndx()
num_parts = 3
ndx_list = []
for i in xrange(num_parts):
for j in xrange(num_parts):
ndx_ij = ndx1.split(i + 1, num_parts, j + 1, num_parts)
ndx_list.append(ndx_ij)
ndx2 = TrialNdx.merge(ndx_list)
assert (ndx1 == ndx2)
def load_scores(score_file, enr_coh_file, coh_test_file, coh_coh_file):
scores = TrialScores.load(score_file)
scores_enr_coh = None
scores_coh_test = None
scores_coh_coh = None
if enr_coh_file is not None:
ndx = TrialNdx(scores.model_set, scores_enr_coh.seg_set)
scores_enr_coh = TrialScores.load(enr_coh_file)
scores_enr_coh = scores_enr_coh.align_with_ndx(ndx)
if coh_test_file is not None:
ndx = TrialNdx(scores_coh_test.model_set, scores.seg_set)
scores_coh_test = TrialScores.load(coh_test_file)
scores_coh_test = scores_coh_test.align_with_ndx(ndx)
if coh_coh_file is not None:
assert scores_enr_coh is not None and scores_coh_test is not None
ndx = TrialNdx(scores_coh_test.model_set, scores_enr_coh.seg_set)
scores_coh_coh = TrialScores.load(coh_coh_file)
scores_coh_coh = scores_coh_coh.align_with_ndx(ndx)
return scores, scores_enr_coh, scores_coh_test, scores_coh_coh
def test_merge():
ndx1 = create_ndx()
ndx2 = TrialNdx(ndx1.model_set[:10], ndx1.seg_set, ndx1.trial_mask[:10, :])
ndx3 = TrialNdx(ndx1.model_set[5:], ndx1.seg_set, ndx1.trial_mask[5:, :])
ndx4 = TrialNdx.merge([ndx2, ndx3])
assert (ndx1 == ndx4)
ndx2 = TrialNdx(ndx1.model_set, ndx1.seg_set[:10], ndx1.trial_mask[:, :10])
ndx3 = TrialNdx(ndx1.model_set, ndx1.seg_set[5:], ndx1.trial_mask[:, 5:])
ndx4 = TrialNdx.merge([ndx2, ndx3])
assert (ndx1 == ndx4)
def eval_calibration(in_score_file, ndx_file, model_file, out_score_file):
logging.info('load ndx: %s' % ndx_file)
try:
ndx = TrialNdx.load_txt(ndx_file)
except:
ndx = TrialKey.load_txt(ndx_file)
logging.info('load scores: %s' % in_score_file)
scr = TrialScores.load_txt(in_score_file)
scr = scr.align_with_ndx(ndx)
logging.info('load model: %s' % model_file)
lr = LR.load(model_file)
logging.info('apply calibration')
s_cal = lr.predict(scr.scores.ravel())
scr.scores = np.reshape(s_cal, scr.scores.shape)
logging.info('save scores: %s' % out_score_file)
scr.save_txt(out_score_file)
def convert(input_file, output_file, test_list, class_file, add_ext):
scores = TrialScores.load(input_file)
if test_list is None:
seg_set = scores.seg_set
else:
with open(test_list, 'r') as f:
seg_set = [
seg for seg in [line.rstrip().split(' ')[0] for line in f]
if seg != 'segmentid'
]
if add_ext:
exts = [os.path.splitext(seg)[1] for seg in seg_set]
seg_set = [os.path.splitext(seg)[0] for seg in seg_set]
if class_file is None:
model_set = scores.model_set
else:
with open(class_file, 'r') as f:
model_set = [line.rstrip().split()[0] for line in f]
ndx = TrialNdx(model_set, seg_set)
scores = scores.set_missing_to_value(ndx, -100)
if add_ext:
scores.seg_set = [seg + ext for seg, ext in zip(scores.seg_set, exts)]
with open(output_file, 'w') as f:
f.write('segmentid\t')
for model in scores.model_set[:-1]:
f.write('%s\t' % model)
f.write('%s\n' % scores.model_set[-1])
for i in xrange(scores.scores.shape[1]):
f.write('%s\t' % scores.seg_set[i])
for j in xrange(scores.scores.shape[0] - 1):
f.write('%f\t' % scores.scores[j, i])
f.write('%f\n' % scores.scores[-1, i])
def create_ndx(ndx_file='./tests/data_in/core-core_det5_ndx.h5'):
ndx = TrialNdx.load(ndx_file)
ndx.sort()
return ndx