def get_param_info_list(nnweight_cfg):
# new way to try and specify config options.
# not sure if i like it yet
param_info_list = ut.flatten([
[
ut.ParamInfo('ratio_thresh', None, type_=float, hideif=None),
ut.ParamInfoBool('lnbnn_on', True, hideif=False),
ut.ParamInfoBool('const_on', False, hideif=False),
ut.ParamInfoBool('borda_on', False, hideif=False),
ut.ParamInfoBool('lograt_on', False, hideif=False),
#ut.ParamInfoBool('loglnbnn_on', False, hideif=False),
#ut.ParamInfoBool('logdist_on', False, hideif=False),
#ut.ParamInfoBool('dist_on', False, hideif=False),
ut.ParamInfoBool('normonly_on', False, hideif=False),
ut.ParamInfoBool('bar_l2_on', False, hideif=False),
ut.ParamInfoBool('cos_on', False, hideif=False),
ut.ParamInfoBool('fg_on', True, hideif=False),
ut.ParamInfo('normalizer_rule', 'last', '', valid_values=['last', 'name']),
ut.ParamInfo('lnbnn_normer', None, hideif=None,
help_='config string for lnbnn score normalizer'),
ut.ParamInfo('lnbnn_norm_thresh', .5, type_=float,
hideif=lambda cfg: not cfg['lnbnn_normer'] ,
help_='config string for lnbnn score normalizer'),
#
ut.ParamInfoBool('can_match_sameimg', False, 'sameimg',
hideif=False),
ut.ParamInfoBool('can_match_samename', True, 'samename',
hideif=True),
# Hacked in
#ut.ParamInfoBool('root_sift_on', False, hideif=False),
ut.ParamInfoBool('sqrd_dist_on', False, hideif=True),
#ut.ParamInfoBool('sqrd_dist_on', True, hideif=True),
],
])
return param_info_list
class VsoneConfig(object):
_param_info_list = [
ut.ParamInfo('distinctiveness_model'),
ut.ParamInfo('score_norm_model'),
]
pass
class ThumbnailConfig(dtool.Config):
_param_info_list = [
ut.ParamInfo('draw_annots', True, hideif=True),
ut.ParamInfo('thumbsize', None, hideif=None),
ut.ParamInfo('ext', '.png', hideif='.png'),
ut.ParamInfo('force_serial', False, hideif=False),
]
def get_param_info_list(self):
return [
ut.ParamInfo('manual_extract', False, hideif=False),
#ut.ParamInfo('ntversion', 1)
ut.ParamInfo('version', 4),
ut.ParamInfo('kp_net', '128_decoupled'),
]
def get_param_info_list(self):
return [
ut.ParamInfo('K', 4),
ut.ParamInfo('Knorm', 1),
ut.ParamInfo('checks', 800),
ut.ParamInfo('version', 1),
]
def get_param_info_list(self):
return [
ut.ParamInfo('csize_max', 8),
ut.ParamInfo('csize_min', 2),
ut.ParamInfo('csize_step', 2),
#ut.ParamInfo('sizes', [2, 4, 6, 8]), # these are percentage (as ints) of trailing edge width
ut.ParamInfo('version', 2),
]
class VocabConfig(dtool.Config):
_param_info_list = [
ut.ParamInfo('algorithm', 'minibatch', 'alg'),
ut.ParamInfo('random_seed', 42, 'seed'),
ut.ParamInfo('num_words', 1000, 'n'),
ut.ParamInfo('version', 2),
ut.ParamInfo('n_init', 1),
]
class FeatConfig(dtool.Config):
_param_info_list = [
ut.ParamInfo('affine_invariance', True),
ut.ParamInfo('rotation_invariance', False),
ut.ParamInfo('augment_orientation', False),
ut.ParamInfo('dense', False),
ut.ParamInfo('dense_stride', False, hideif=lambda cfg: not cfg['dense']),
]
def get_param_info_list(self):
return [
ut.ParamInfo('crop_dim_size', 750, 'sz', hideif=750),
ut.ParamInfo('crop_enabled', True, hideif=False),
#ut.ParamInfo('ccversion', 1)
ut.ParamInfo('version', 2),
ut.ParamInfo('ext', '.png'),
]
def get_param_info_list(occur_cfg):
param_info_list = [
ut.ParamInfo('min_imgs_per_occurrence', 1, 'minper='),
ut.ParamInfo('cluster_algo', 'agglomerative', '', valid_values=['agglomerative', 'meanshift']),
ut.ParamInfo('quantile', .01, 'quant', hideif=lambda cfg: cfg['cluster_algo'] != 'meanshift'),
ut.ParamInfo('seconds_thresh', 600, 'sec', hideif=lambda cfg: cfg['cluster_algo'] != 'agglomerative'),
ut.ParamInfo('use_gps', False, hideif=False),
]
return param_info_list
class InferenceConfig(dtool.Config):
_param_info_list = [
ut.ParamInfo('min_labels', 1),
ut.ParamInfo('max_labels', 5),
ut.ParamInfo('thresh_method',
'elbow',
valid_values=['elbow', 'mean', 'median', 'custom']),
ut.ParamInfo('thresh',
.5,
hideif=lambda cfg: cfg['thresh_method'] != 'custom'),
]
def get_param_info_list(self):
return [
#ut.ParamInfo('score_method', 'csum'),
# should this be the only thing here?
#ut.ParamInfo('daids', None),
ut.ParamInfo('decision', 'max'),
#ut.ParamInfo('sizes', (5, 10, 15, 20)),
ut.ParamInfo('weight_import', 1),
ut.ParamInfo('window', 10),
#ut.ParamInfo('bcdtwversion', 1),
ut.ParamInfo('version', 8),
]
class InvertedIndexConfig(dtool_ibeis.Config):
_param_info_list = [
ut.ParamInfo('nAssign', 1),
#ut.ParamInfo('int_rvec', False, hideif=False),
ut.ParamInfo('int_rvec', True, hideif=False),
ut.ParamInfo('massign_equal,', False),
ut.ParamInfo('massign_alpha,', 1.2),
# ut.ParamInfo('massign_sigma,', 80.0, hideif=lambda cfg: cfg['massign_equal']),
ut.ParamInfo('inva_version', 2),
#
#massign_sigma=80.0,
#massign_equal_weights=False
]
class ClassifierConfig(dtool.Config):
_param_info_list = [
ut.ParamInfo('classifier_sensitivity', None),
]
_sub_config_list = [
ThumbnailConfig
]
def rectify_item(key, val):
if val is None:
return ut.ParamInfo(key, val)
elif isinstance(val, ut.ParamInfo):
if val.varname is None:
# Copy and assign a new varname
pi = copy.deepcopy(val)
pi.varname = key
else:
pi = val
assert pi.varname == key, (
'Given varname=%r does not match key=%r' %
(pi.varname, key))
return pi
else:
return ut.ParamInfo(key, val, type_=type(val))
class PairFeatureConfig(dt.Config):
"""
Config for building pairwise feature dimensions
I.E. Config to distil unordered feature correspondences into a fixed length
vector.
"""
_param_info_list = [
# ut.ParamInfo('indices', slice(0, 5)),
ut.ParamInfo('indices', []),
ut.ParamInfo('summary_ops', {
# 'invsum',
'sum', 'std', 'mean', 'len', 'med'}),
ut.ParamInfo('local_keys', None),
ut.ParamInfo('sorters', [
# 'ratio', 'norm_dist', 'match_dist'
# 'lnbnn', 'lnbnn_norm_dist',
]),
# ut.ParamInfo('bin_key', None, valid_values=[None, 'ratio']),
ut.ParamInfo('bin_key', 'ratio', valid_values=[None, 'ratio']),
# ut.ParamInfo('bins', [.5, .6, .7, .8])
# ut.ParamInfo('bins', None, type_=eval),
ut.ParamInfo('bins', (.625,), type_=eval),
# ut.ParamInfo('need_lnbnn', False),
ut.ParamInfo('use_na', False), # change to True if sklearn has RFs with nan support
]
class VocabConfig(dtool.Config):
"""
Example:
>>> from ibeis.core_annots import * # NOQA
>>> cfg = VocabConfig()
>>> result = str(cfg)
>>> print(result)
"""
_param_info_list = [
ut.ParamInfo('algorithm', 'kdtree', 'alg'),
ut.ParamInfo('random_seed', 42, 'seed'),
ut.ParamInfo('num_words', 1000, 'seed'),
ut.ParamInfo('version', 1),
# max iters
# flann params
# random seed
]
_sub_config_list = []
class DummyVsManyConfig(dtool.Config):
# Different pipeline components can go here as well as dependencies
# that were not explicitly enumerated in the tree structure
_param_info_list = [
#ut.ParamInfo('score_method', 'csum'),
# should this be the only thing here?
#ut.ParamInfo('daids', None),
ut.ParamInfo('distinctiveness_model', None),
ut.ParamInfo('version', 2),
]
_sub_config_list = [
# I guess different annots might want different configs ...
DummyChipConfig,
DummyKptsConfig,
DummyIndexerConfig,
DummyNNConfig,
DummySVERConfig
]
def rectify_item(key, val):
if val is None:
return ut.ParamInfo(key, val)
elif isinstance(val, ut.ParamInfo):
if val.varname is None:
# Copy and assign a new varname
pi = copy.deepcopy(val)
pi.varname = key
else:
pi = val
assert pi.varname == key, (
'Given varname=%r does not match key=%r' % (pi.varname, key))
return pi
else:
if isinstance(val, Config):
# Set table name from key when doing nested from dicts
if val.__class__.__name__ == 'UnnamedConfig':
val.__class__.__name__ = key + 'Config'
return ut.ParamInfo(key, val, type_=type(val))
def gridsearch_addWeighted():
r"""
CommandLine:
python -m vtool.blend --test-gridsearch_addWeighted --show
Example:
>>> # GRIDSEARCH
>>> from vtool.blend import * # NOQA
>>> gridsearch_addWeighted()
>>> ut.show_if_requested()
"""
import cv2
import vtool as vt
def test_func(src1, src2, alpha=1.0, **kwargs):
beta = 1.0 - alpha
src1 = vt.rectify_to_float01(src1)
src2 = vt.rectify_to_float01(src2)
dst = np.empty(src1.shape, dtype=src1.dtype)
cv2.addWeighted(src1=src1,
src2=src2,
dst=dst,
alpha=alpha,
beta=beta,
dtype=-1,
**kwargs)
return dst
img1, img2 = testdata_blend()
args = img1, img2 = vt.make_channels_comparable(img1, img2)
param_info = ut.ParamInfoList(
'blend_params',
[
ut.ParamInfo('alpha', .8, varyvals=np.linspace(0, 1.0,
5).tolist()),
#ut.ParamInfo('beta', .8,
ut.ParamInfo('gamma', .0, varyvals=np.linspace(0, 1.0,
5).tolist()),
#varyvals=[.0],))
#ut.ParamInfo('gamma', .8, 'alpha=',
# varyvals=np.linspace(0, 1.0, 9).tolist()),
])
gridsearch_image_function(param_info, test_func, args)
def get_param_info_list(self):
return [
ut.ParamInfo('n_neighbors', 3, 'n_nb'),
ut.ParamInfo('ignore_notch', True, 'ign_n', hideif=False),
#ut.ParamInfo('teversion', 1),
ut.ParamInfo('version', 9),
ut.ParamInfo('use_te_scorer', True, 'te_s', hideif=False),
ut.ParamInfo('te_score_weight', 0.5, 'w_tes'),
ut.ParamInfo('te_net', 'annot_res'),
ut.ParamInfo('te_score_method', 'avg', 'te_sm'),
ut.ParamInfo(
'tol', None
), # allow the trailing edge to go x percentage of the image height below
]
class ProbchipConfig(dtool.Config):
"""
CommandLine:
python -m dtool.example_depcache --exec-ProbchipConfig --show
Example:
>>> # DISABLE_DOCTEST
>>> from wbia.dtool.depcache_control import * # NOQA
>>> from wbia.dtool.example_depcache import testdata_depc
>>> depc = testdata_depc()
>>> table = depc['probchip']
>>> exec(ut.execstr_funckw(table.get_rowid), globals())
>>> config = table.configclass(testerror=True)
>>> root_rowids = [1, 2, 3]
>>> parent_rowids = list(zip(root_rowids))
>>> proptup_gen = list(table.preproc_func(depc, root_rowids, config))
>>> pc_rowids = depc.get_rowids('probchip', root_rowids, config)
>>> prop_list2 = depc.get('probchip', root_rowids, config=config, read_extern=False)
>>> print(prop_list2)
>>> #depc.new_request('probchip', [1, 2, 3])
>>> fg_rowids = depc.get_rowids('fgweight', root_rowids, config)
>>> fg = depc.get('fgweight', root_rowids, config=config)
>>> #############
>>> config = table.configclass(testerror=False)
>>> root_rowids = [1, 2, 3]
>>> parent_rowids = list(zip(root_rowids))
>>> proptup_gen = list(table.preproc_func(depc, root_rowids, config))
>>> pc_rowids2 = depc.get_rowids('probchip', root_rowids, config)
>>> prop_list2 = depc.get('probchip', root_rowids, config=config, read_extern=False)
>>> print(prop_list2)
>>> #depc.new_request('probchip', [1, 2, 3])
>>> fg_rowids2 = depc.get_rowids('fgweight', root_rowids, config)
"""
_param_info_list = [
ut.ParamInfo('testerror', False, hideif=False),
ut.ParamInfo('ext', '.png', hideif='.png'),
]
class NormFeatScoreConfig(dtool.Config):
_alias = 'nfscfg'
_param_info_list = [
ut.ParamInfo('disttype', None),
ut.ParamInfo('namemode', True),
ut.ParamInfo('fsvx', None, type_='fuzzy_subset', hideif=None),
ut.ParamInfo('threshx', None, hideif=None),
ut.ParamInfo('thresh', .9, hideif=lambda cfg: cfg['threshx'] is None),
ut.ParamInfo('num', 5),
# ut.ParamInfo('top_percent', None, hideif=None),
ut.ParamInfo('top_percent', .5, hideif=None),
]
class LocalizerConfig(dtool.Config):
_param_info_list = [
ut.ParamInfo('algo', 'yolo'),
ut.ParamInfo('sensitivity', 0.0),
ut.ParamInfo('species', 'zebra_plains', hideif='zebra_plains'),
ut.ParamInfo('config_filepath', None),
ut.ParamInfo('weight_filepath', None),
ut.ParamInfo('grid', False),
]
_sub_config_list = [
ThumbnailConfig
]
class DetectorConfig(dtool.Config):
_param_info_list = [
ut.ParamInfo('classifier_sensitivity', 0.82),
ut.ParamInfo('localizer_config_filepath', None),
ut.ParamInfo('localizer_weight_filepath', None),
ut.ParamInfo('localizer_grid', False),
ut.ParamInfo('localizer_sensitivity', 0.16),
ut.ParamInfo('labeler_sensitivity', 0.42),
]
_sub_config_list = [
ThumbnailConfig,
LocalizerConfig,
]
def get_param_info_list(rrvsone_cfg):
# new way to try and specify config options.
# not sure if i like it yet
param_info_list = ut.flatten([
[
ut.ParamInfo('index_method', 'single', ''),
ut.ParamInfo('K', 4, type_=int),
ut.ParamInfo('Knorm', 1, 'Kn='),
ut.ParamInfo('use_k_padding', False, 'padk='),
ut.ParamInfo('single_name_condition', False, 'nameknn', type_=bool, hideif=False),
ut.ParamInfo('checks', 800, 'cks', type_=int),
#ut.ParamInfo('ratio_thresh', None, type_=float, hideif=None),
],
])
return param_info_list
def get_param_info_list(rrvsone_cfg):
from ibeis.algo.hots import distinctiveness_normalizer
from ibeis.algo.hots import vsone_pipeline
# new way to try and specify config options.
# not sure if i like it yet
param_info_list = ut.flatten([
[
ut.ParamInfo('rrvsone_on', False, ''),
],
vsone_pipeline.OTHER_RRVSONE_PARAMS.aslist(),
vsone_pipeline.SHORTLIST_DEFAULTS.aslist(),
vsone_pipeline.COEFF_DEFAULTS.aslist(),
vsone_pipeline.UNC_DEFAULTS.aslist(),
vsone_pipeline.SCR_DEFAULTS.aslist(),
vsone_pipeline.COVKPTS_DEFAULT.aslist(
hideif=lambda cfg: not cfg['covscore_on'] or cfg['maskscore_mode'] != 'kpts'),
vsone_pipeline.COVGRID_DEFAULT.aslist(
hideif=lambda cfg: not cfg['covscore_on'] or cfg['maskscore_mode'] != 'grid'),
distinctiveness_normalizer.DCVS_DEFAULT.aslist(
hideif=lambda cfg: not cfg['dcvs_on']),
])
return param_info_list
class SMKRequestConfig(dtool_ibeis.Config):
""" Figure out how to do this """
_param_info_list = [
ut.ParamInfo('proot', 'smk'),
ut.ParamInfo('smk_alpha', 3.0),
ut.ParamInfo('smk_thresh', 0.0),
#ut.ParamInfo('smk_thresh', -1.0),
ut.ParamInfo('agg', True),
ut.ParamInfo('data_ma',
False), # hack for query only multiple assignment
ut.ParamInfo(
'word_weight_method', 'idf',
shortprefix='wwm'), # hack for query only multiple assignment
ut.ParamInfo('smk_version', 3),
]
_sub_config_list = [
core_annots.ChipConfig,
core_annots.FeatConfig,
old_config.SpatialVerifyConfig,
vocab_indexer.VocabConfig,
inverted_index.InvertedIndexConfig,
MatchHeuristicsConfig,
]
class DummyChipConfig(dtool.Config):
"""
Example:
>>> # ENABLE_DOCTEST
>>> from dtool.example_depcache import * # NOQA
>>> cfg = DummyChipConfig()
>>> cfg.dim_size = 700
>>> cfg.histeq = True
>>> print(cfg)
>>> cfg.histeq = False
>>> print(cfg)
"""
_param_info_list = [
ut.ParamInfo('resize_dim',
'width',
valid_values=['area', 'width', 'heigh', 'diag']),
ut.ParamInfo('dim_size', 500, 'sz'),
ut.ParamInfo('preserve_aspect', True),
ut.ParamInfo('histeq', False, hideif=False),
ut.ParamInfo('ext', '.png'),
ut.ParamInfo('version', 0),
]
class MatchHeuristicsConfig(dtool.Config):
_param_info_list = [
ut.ParamInfo('can_match_self', False),
ut.ParamInfo('can_match_samename', True),
ut.ParamInfo('can_match_sameimg', False),
]
