def get_infostr(qreq_):
infostr_list = []
app = infostr_list.append
qaid_internal = qreq_.get_internal_qaids()
daid_internal = qreq_.get_internal_daids()
qd_intersection = ut.intersect_ordered(daid_internal, qaid_internal)
app(' * len(internal_daids) = %r' % len(daid_internal))
app(' * len(internal_qaids) = %r' % len(qaid_internal))
app(' * len(qd_intersection) = %r' % len(qd_intersection))
infostr = '\n'.join(infostr_list)
return infostr
def get_infostr(qreq_):
infostr_list = []
app = infostr_list.append
qaid_internal = qreq_.get_internal_qaids()
daid_internal = qreq_.get_internal_daids()
qd_intersection = ut.intersect_ordered(daid_internal, qaid_internal)
app(' * len(internal_daids) = %r' % len(daid_internal))
app(' * len(internal_qaids) = %r' % len(qaid_internal))
app(' * len(qd_intersection) = %r' % len(qd_intersection))
infostr = '\n'.join(infostr_list)
return infostr
def choose_vsmany_K(num_names, qaids, daids):
"""
TODO: Should also scale up the number of checks as well
method for choosing K in the initial vsmany queries
Ignore:
>>> # DISABLE_DOCTEST
>>> # Shows plot for K vs number of names
>>> from ibeis.algo.hots.automated_params import * # NOQA
>>> import ibeis
>>> from ibeis import constants as const
>>> ibs = ibeis.opendb(defaultdb='testdb1')
>>> valid_aids = ibs.get_valid_aids(species=const.TEST_SPECIES.ZEB_PLAIN)
>>> num_names = np.arange(0, 1000)
>>> num_names_slope = .1
>>> K_max = 10
>>> K_min = 1
>>> K_list = np.floor(num_names_slope * num_names)
>>> K_list[K_list > K_max] = K_max
>>> K_list[K_list < K_min] = K_min
>>> clip_index_list = np.where(K_list >= K_max)[0]
>>> clip_index = clip_index_list[min(len(clip_index_list) - 1, 10)]
>>> K_list = K_list[0:clip_index]
>>> num_names = num_names[0:clip_index]
>>> pt.plot2(num_names, K_list, x_label='num_names', y_label='K',
... equal_aspect=False, marker='g-', pad=1, dark=True)
>>> pt.update()
"""
#K = ibs.cfg.query_cfg.nn_cfg.K
# TODO: paramaterize in config
num_names_slope = .1 # increase K every fifty names
K_max = 10
K_min = 1
num_names_lower = K_min / num_names_slope
num_names_upper = K_max / num_names_slope
if num_names < num_names_lower:
K = K_min
elif num_names < num_names_upper:
K = num_names_slope * num_names
else:
K = K_max
with ut.embed_on_exception_context:
if len(ut.intersect_ordered(qaids, daids)) > 0:
# if self is in query bump k
K += 1
return K
def choose_vsmany_K(num_names, qaids, daids):
"""
TODO: Should also scale up the number of checks as well
method for choosing K in the initial vsmany queries
Ignore:
>>> # DISABLE_DOCTEST
>>> # Shows plot for K vs number of names
>>> from ibeis.algo.hots.automated_params import * # NOQA
>>> import ibeis
>>> from ibeis import constants as const
>>> ibs = ibeis.opendb(defaultdb='testdb1')
>>> valid_aids = ibs.get_valid_aids(species=const.TEST_SPECIES.ZEB_PLAIN)
>>> num_names = np.arange(0, 1000)
>>> num_names_slope = .1
>>> K_max = 10
>>> K_min = 1
>>> K_list = np.floor(num_names_slope * num_names)
>>> K_list[K_list > K_max] = K_max
>>> K_list[K_list < K_min] = K_min
>>> clip_index_list = np.where(K_list >= K_max)[0]
>>> clip_index = clip_index_list[min(len(clip_index_list) - 1, 10)]
>>> K_list = K_list[0:clip_index]
>>> num_names = num_names[0:clip_index]
>>> pt.plot2(num_names, K_list, x_label='num_names', y_label='K',
... equal_aspect=False, marker='g-', pad=1, dark=True)
>>> pt.update()
"""
#K = ibs.cfg.query_cfg.nn_cfg.K
# TODO: paramaterize in config
num_names_slope = .1 # increase K every fifty names
K_max = 10
K_min = 1
num_names_lower = K_min / num_names_slope
num_names_upper = K_max / num_names_slope
if num_names < num_names_lower:
K = K_min
elif num_names < num_names_upper:
K = num_names_slope * num_names
else:
K = K_max
with ut.embed_on_exception_context:
if len(ut.intersect_ordered(qaids, daids)) > 0:
# if self is in query bump k
K += 1
return K
