def viz_factor_graph(gm):
"""
ut.qtensure()
gm = build_factor_graph(G, nodes, edges , n_annots, n_names, lookup_annot_idx,
use_unaries=True, edge_probs=None, operator='multiplier')
"""
ut.qtensure()
import networkx
from networkx.drawing.nx_agraph import graphviz_layout
networkx.graphviz_layout = graphviz_layout
opengm.visualizeGm(gm, show=False, layout="neato", plotUnaries=True,
iterations=1000, plotFunctions=True,
plotNonShared=False, relNodeSize=1.0)
_ = pt.show_nx(gm.G) # NOQA
def compare_data(Y_list_):
import wbia
qreq_ = wbia.testdata_qreq_(
defaultdb='Oxford',
a='oxford',
p=
'smk:nWords=[64000],nAssign=[1],SV=[False],can_match_sameimg=True,dim_size=None',
)
qreq_.ensure_data()
gamma1s = []
gamma2s = []
logger.info(len(Y_list_))
logger.info(len(qreq_.daids))
dinva = qreq_.dinva
bady = []
for Y in Y_list_:
aid = Y.aid
gamma1 = Y.gamma
if aid in dinva.aid_to_idx:
idx = dinva.aid_to_idx[aid]
gamma2 = dinva.gamma_list[idx]
gamma1s.append(gamma1)
gamma2s.append(gamma2)
else:
bady += [Y]
logger.info(Y.nid)
# logger.info(Y.qual)
# ibs = qreq_.ibs
# z = ibs.annots([a.aid for a in bady])
import wbia.plottool as pt
ut.qtensure()
gamma1s = np.array(gamma1s)
gamma2s = np.array(gamma2s)
sortx = gamma1s.argsort()
pt.plot(gamma1s[sortx], label='script')
pt.plot(gamma2s[sortx], label='pipe')
pt.legend()
def sift_test():
"""
Play with SIFT equations using python so I can see and compare results.
"""
import numpy as np
# Sample measurement from lena
sift_raw = np.array([
48.0168,
130.017,
159.065,
54.5727,
63.7103,
14.3629,
27.0228,
15.3527,
40.5067,
165.721,
511.036,
196.888,
4.72748,
8.85093,
15.9457,
14.4198,
49.7571,
209.104,
452.047,
223.972,
2.66391,
16.8975,
21.7488,
13.6855,
0.700244,
10.2518,
312.483,
282.647,
1.82898,
3.01759,
0.448028,
0,
144.834,
300.438,
131.837,
40.3284,
11.1998,
9.68647,
7.68484,
29.166,
425.953,
386.903,
352.388,
267.883,
12.9652,
18.833,
8.55462,
71.7924,
112.282,
295.512,
678.599,
419.405,
21.3151,
91.9408,
22.8681,
9.83749,
3.06347,
97.6562,
458.799,
221.873,
68.1473,
410.764,
48.9493,
2.01682,
194.794,
43.7171,
16.2078,
17.5604,
48.8504,
48.3823,
45.7636,
299.432,
901.565,
188.732,
32.6512,
23.6874,
55.379,
272.264,
68.2334,
221.37,
159.631,
44.1475,
126.636,
95.1978,
74.1097,
1353.24,
239.319,
33.5368,
5.62254,
69.0013,
51.7629,
9.55458,
26.4599,
699.623,
208.78,
2.09156,
135.278,
19.5378,
52.0265,
51.8445,
49.1938,
9.04161,
11.6605,
87.4498,
604.012,
85.6801,
42.9738,
75.8549,
183.65,
206.912,
34.2781,
95.0146,
13.4201,
83.7426,
440.322,
83.0038,
125.663,
457.333,
52.6424,
4.93713,
0.38947,
244.762,
291.113,
7.50165,
8.16208,
73.2169,
21.9674,
0.00429259,
])
import vtool as vt
# CONFIRMED: One normalization followed by another does not do anything
#sift_root1 = vt.normalize(sift_root1, ord=2)
#sift_root1 = vt.normalize(sift_root1, ord=1)
sift_clip = sift_raw.copy()
sift_clip = vt.normalize(sift_clip, ord=2)
sift_clip[sift_clip > .2] = .2
sift_clip = vt.normalize(sift_clip, ord=2)
siff_ell2 = vt.normalize(sift_raw, ord=2)
siff_ell1 = vt.normalize(sift_raw, ord=1)
# Two versions of root SIFT
# They are equlivalent
# taken from https://hal.inria.fr/hal-00840721/PDF/RR-8325.pdf
normalize1 = lambda x: vt.normalize(x, ord=1) # NOQA
normalize2 = lambda x: vt.normalize(x, ord=2) # NOQA
assert np.all(
np.isclose(np.sqrt(normalize1(sift_raw)),
normalize2(np.sqrt(sift_raw))))
# How do we genralize this for alpha != .5?
# Just always L2 normalize afterwords?
alpha = .2
powerlaw = lambda x: np.power(x, alpha) # NOQA
sift_root1 = normalize2(powerlaw(normalize1(sift_raw)))
sift_root2 = normalize2(powerlaw(sift_raw))
flags = np.isclose(sift_root1, sift_root2)
print(flags)
assert np.all(flags)
#sift_root_quant = np.clip((sift_root1 * 512), 0, 255).astype(np.uint8)
#p = (np.bincount(sift_root_quant) / 128)
#entropy = -np.nansum(p * np.log2(p))
s = sift_raw[0:10]
np.sqrt(s) / (np.sqrt(s).sum()**2)
np.power(normalize1(s), 2)
#b = powerlaw(normalize1(s))
#print(np.isclose(a, b))
np.isclose(normalize1(s), normalize1(normalize2(s)))
# Another root SIFT version from
# https://hal.inria.fr/hal-00688169/document
# but this doesnt seem to work with uint8 representations
sift_root3 = np.sqrt(sift_raw)
sift_root3 = sift_root3 / np.sqrt(np.linalg.norm(sift_root3))
import plottool as pt
import utool as ut
ut.qtensure()
fig = pt.figure(fnum=1, pnum=None)
def draw_sift(sift, pnum, title, **kwargs):
ax = fig.add_subplot(*pnum)
pt.draw_sifts(ax, sift[None, :], **kwargs)
ax.set_xlim(-1, 1)
ax.set_ylim(-1, 1)
ax.grid(False)
ax.set_aspect('equal')
ax.set_xticks([])
ax.set_yticks([])
if title:
ax.set_title(title)
fig.clf()
pnum_ = pt.make_pnum_nextgen(2, 4)
draw_sift(sift_raw, pnum_(), 'raw/max(raw)', fidelity=sift_raw.max())
draw_sift(sift_clip, pnum_(), 'clip', fidelity=1.0)
draw_sift(siff_ell2, pnum_(), 'l2', fidelity=1.0)
draw_sift(siff_ell1, pnum_(), 'l1', fidelity=1.0)
draw_sift(sift_root1, pnum_(), 'root1', fidelity=1.0)
draw_sift(sift_root2, pnum_(), 'root2', fidelity=1.0)
draw_sift(sift_root3, pnum_(), 'root3', fidelity=2.0)
def fix_annotmatch_pzmaster1():
"""
PZ_Master1 had annotmatch rowids that did not agree with the current name
labeling. Looking at the inconsistencies in the graph interface was too
cumbersome, because over 3000 annots were incorrectly grouped together.
This function deletes any annotmatch rowid that is not consistent with the
current labeling so we can go forward with using the new AnnotInference
object
"""
import wbia
ibs = wbia.opendb('PZ_Master1')
infr = wbia.AnnotInference(ibs=ibs, aids=ibs.get_valid_aids(), verbose=5)
infr.initialize_graph()
annots = ibs.annots()
aid_to_nid = ut.dzip(annots.aids, annots.nids)
if False:
infr.reset_feedback()
infr.ensure_mst()
infr.apply_feedback_edges()
infr.relabel_using_reviews()
infr.start_qt_interface()
# Get annotmatch rowids that agree with current labeling
if False:
annotmatch = ibs.db.get_table_as_pandas('annotmatch')
import pandas as pd
flags1 = pd.isnull(annotmatch['annotmatch_evidence_decision'])
flags2 = annotmatch['annotmatch_tag_text'] == ''
bad_part = annotmatch[flags1 & flags2]
rowids = bad_part.index.tolist()
ibs.delete_annotmatch(rowids)
if False:
# Delete bidirectional annotmatches
annotmatch = ibs.db.get_table_as_pandas('annotmatch')
df = annotmatch.set_index(['annot_rowid1', 'annot_rowid2'])
# Find entires that have both directions
pairs1 = annotmatch[['annot_rowid1', 'annot_rowid2']].values
f_edges = {tuple(p) for p in pairs1}
b_edges = {tuple(p[::-1]) for p in pairs1}
isect_edges = {tuple(sorted(p)) for p in b_edges.intersection(f_edges)}
isect_edges1 = list(isect_edges)
isect_edges2 = [p[::-1] for p in isect_edges]
# cols = ['annotmatch_evidence_decision', 'annotmatch_tag_text']
import pandas as pd
custom_ = {
(559, 4909): (False, ['photobomb']),
(7918, 8041): (False, ['photobomb']),
(6634, 6754): (False, ['photobomb']),
(3707, 3727): (False, ['photobomb']),
(86, 103): (False, ['photobomb']),
}
extra_ = {}
fixme_edges = []
d1 = df.loc[isect_edges1].reset_index(drop=False)
d2 = df.loc[isect_edges2].reset_index(drop=False)
flags = d1['annotmatch_evidence_decision'] != d2[
'annotmatch_evidence_decision']
from wbia.tag_funcs import _parse_tags
for f, r1, r2 in zip(flags, d1.iterrows(), d2.iterrows()):
v1, v2 = r1[1], r2[1]
aid1 = v1['annot_rowid1']
aid2 = v1['annot_rowid2']
truth_real = (ibs.const.EVIDENCE_DECISION.POSITIVE
if aid_to_nid[aid1] == aid_to_nid[aid2] else
ibs.const.EVIDENCE_DECISION.NEGATIVE)
truth1 = v1['annotmatch_evidence_decision']
truth2 = v2['annotmatch_evidence_decision']
t1 = _parse_tags(v1['annotmatch_tag_text'])
t2 = _parse_tags(v2['annotmatch_tag_text'])
newtag = ut.union_ordered(t1, t2)
if (aid1, aid2) in custom_:
continue
fixme_flag = False
if not pd.isnull(truth1):
if truth_real != truth1:
fixme_flag = True
if not pd.isnull(truth2):
if truth_real != truth2:
fixme_flag = True
if fixme_flag:
logger.info('newtag = %r' % (newtag, ))
logger.info('truth_real = %r' % (truth_real, ))
logger.info('truth1 = %r' % (truth1, ))
logger.info('truth2 = %r' % (truth2, ))
logger.info('aid1 = %r' % (aid1, ))
logger.info('aid2 = %r' % (aid2, ))
fixme_edges.append((aid1, aid2))
else:
extra_[(aid1, aid2)] = (truth_real, newtag)
extra_.update(custom_)
new_pairs = extra_.keys()
new_truths = ut.take_column(ut.dict_take(extra_, new_pairs), 0)
new_tags = ut.take_column(ut.dict_take(extra_, new_pairs), 1)
new_tag_texts = [';'.join(t) for t in new_tags]
aids1, aids2 = ut.listT(new_pairs)
# Delete the old
ibs.delete_annotmatch((d1['annotmatch_rowid'].values.tolist() +
d2['annotmatch_rowid'].values.tolist()))
# Add the new
ams = ibs.add_annotmatch_undirected(aids1, aids2)
ibs.set_annotmatch_evidence_decision(ams, new_truths)
ibs.set_annotmatch_tag_text(ams, new_tag_texts)
if False:
import wbia.guitool as gt
gt.ensure_qapp()
ut.qtensure()
from wbia.gui import inspect_gui
inspect_gui.show_vsone_tuner(ibs, aid1, aid2)
# pairs2 = pairs1.T[::-1].T
# idx1, idx2 = ut.isect_indices(list(map(tuple, pairs1)),
# list(map(tuple, pairs2)))
# r_edges = list(set(map(tuple, map(sorted, pairs1[idx1]))))
# unique_pairs = list(set(map(tuple, map(sorted, pairs1[idx1]))))
# df = annotmatch.set_index(['annot_rowid1', 'annot_rowid2'])
x = ut.ddict(list)
annotmatch = ibs.db.get_table_as_pandas('annotmatch')
import ubelt as ub
_iter = annotmatch.iterrows()
prog = ub.ProgIter(_iter, length=len(annotmatch))
for k, m in prog:
aid1 = m['annot_rowid1']
aid2 = m['annot_rowid2']
if m['annotmatch_evidence_decision'] == ibs.const.EVIDENCE_DECISION.POSITIVE:
if aid_to_nid[aid1] == aid_to_nid[aid2]:
x['agree1'].append(k)
else:
x['disagree1'].append(k)
elif m['annotmatch_evidence_decision'] == ibs.const.EVIDENCE_DECISION.NEGATIVE:
if aid_to_nid[aid1] == aid_to_nid[aid2]:
x['disagree2'].append(k)
else:
x['agree2'].append(k)
ub.map_vals(len, x)
ut.dict_hist(annotmatch.loc[x['disagree1']]['annotmatch_tag_text'])
disagree1 = annotmatch.loc[x['disagree1']]
pb_disagree1 = disagree1[disagree1['annotmatch_tag_text'] == 'photobomb']
aids1 = pb_disagree1['annot_rowid1'].values.tolist()
aids2 = pb_disagree1['annot_rowid2'].values.tolist()
aid_pairs = list(zip(aids1, aids2))
infr = wbia.AnnotInference.from_pairs(aid_pairs, ibs=ibs, verbose=5)
if False:
feedback = infr.read_wbia_annotmatch_feedback(edges=infr.edges())
infr.external_feedback = feedback
infr.apply_feedback_edges()
infr.start_qt_interface(loop=False)
# Delete these values
if False:
nonpb_disagree1 = disagree1[
disagree1['annotmatch_tag_text'] != 'photobomb']
disagree2 = annotmatch.loc[x['disagree2']]
ibs.delete_annotmatch(nonpb_disagree1['annotmatch_rowid'])
ibs.delete_annotmatch(disagree2['annotmatch_rowid'])
# ut.dict_hist(disagree1['annotmatch_tag_text'])
import networkx as nx
graph = nx.Graph()
graph.add_edges_from(
zip(pb_disagree1['annot_rowid1'], pb_disagree1['annot_rowid2']))
list(nx.connected_components(graph))
set(annotmatch.loc[x['disagree2']]['annotmatch_tag_text'])
def fix_bidirectional_annotmatch(ibs):
import wbia
infr = wbia.AnnotInference(ibs=ibs, aids='all', verbose=5)
infr.initialize_graph()
annots = ibs.annots()
aid_to_nid = ut.dzip(annots.aids, annots.nids)
# Delete bidirectional annotmatches
annotmatch = ibs.db.get_table_as_pandas('annotmatch')
df = annotmatch.set_index(['annot_rowid1', 'annot_rowid2'])
# Find entires that have both directions
pairs1 = annotmatch[['annot_rowid1', 'annot_rowid2']].values
f_edges = {tuple(p) for p in pairs1}
b_edges = {tuple(p[::-1]) for p in pairs1}
isect_edges = {tuple(sorted(p)) for p in b_edges.intersection(f_edges)}
logger.info('Found %d bidirectional edges' % len(isect_edges))
isect_edges1 = list(isect_edges)
isect_edges2 = [p[::-1] for p in isect_edges]
import pandas as pd
extra_ = {}
fixme_edges = []
d1 = df.loc[isect_edges1].reset_index(drop=False)
d2 = df.loc[isect_edges2].reset_index(drop=False)
flags = d1['annotmatch_evidence_decision'] != d2[
'annotmatch_evidence_decision']
from wbia.tag_funcs import _parse_tags
for f, r1, r2 in zip(flags, d1.iterrows(), d2.iterrows()):
v1, v2 = r1[1], r2[1]
aid1 = v1['annot_rowid1']
aid2 = v1['annot_rowid2']
truth_real = (ibs.const.EVIDENCE_DECISION.POSITIVE
if aid_to_nid[aid1] == aid_to_nid[aid2] else
ibs.const.EVIDENCE_DECISION.NEGATIVE)
truth1 = v1['annotmatch_evidence_decision']
truth2 = v2['annotmatch_evidence_decision']
t1 = _parse_tags(v1['annotmatch_tag_text'])
t2 = _parse_tags(v2['annotmatch_tag_text'])
newtag = ut.union_ordered(t1, t2)
fixme_flag = False
if not pd.isnull(truth1):
if truth_real != truth1:
fixme_flag = True
if not pd.isnull(truth2):
if truth_real != truth2:
fixme_flag = True
if fixme_flag:
logger.info('--')
logger.info('t1, t2 = %r, %r' % (t1, t2))
logger.info('newtag = %r' % (newtag, ))
logger.info('truth_real, truth1, truth2 = %r, %r, %r' %
(truth_real, truth1, truth2))
logger.info('aid1, aid2 = %r, %r' % (aid1, aid2))
fixme_edges.append(tuple(sorted((aid1, aid2))))
else:
extra_[(aid1, aid2)] = (truth_real, newtag)
if len(fixme_edges) > 0:
# need to manually fix these edges
fix_infr = wbia.AnnotInference.from_pairs(fixme_edges,
ibs=ibs,
verbose=5)
feedback = fix_infr.read_wbia_annotmatch_feedback(
only_existing_edges=True)
infr = fix_infr
fix_infr.external_feedback = feedback
fix_infr.apply_feedback_edges()
fix_infr.start_qt_interface(loop=False)
# DELETE OLD EDGES TWICE
ams = ibs.get_annotmatch_rowid_from_edges(fixme_edges)
ibs.delete_annotmatch(ams)
ams = ibs.get_annotmatch_rowid_from_edges(fixme_edges)
ibs.delete_annotmatch(ams)
# MANUALLY CALL THIS ONCE FINISHED
# TO ONLY CHANGE ANNOTMATCH EDGES
infr.write_wbia_staging_feedback()
infr.write_wbia_annotmatch_feedback()
# extra_.update(custom_)
new_pairs = extra_.keys()
new_truths = ut.take_column(ut.dict_take(extra_, new_pairs), 0)
new_tags = ut.take_column(ut.dict_take(extra_, new_pairs), 1)
new_tag_texts = [';'.join(t) for t in new_tags]
aids1, aids2 = ut.listT(new_pairs)
# Delete the old
ibs.delete_annotmatch((d1['annotmatch_rowid'].values.tolist() +
d2['annotmatch_rowid'].values.tolist()))
# Add the new
ams = ibs.add_annotmatch_undirected(aids1, aids2)
ibs.set_annotmatch_evidence_decision(ams, new_truths)
ibs.set_annotmatch_tag_text(ams, new_tag_texts)
if False:
import wbia.guitool as gt
gt.ensure_qapp()
ut.qtensure()
from wbia.gui import inspect_gui
inspect_gui.show_vsone_tuner(ibs, aid1, aid2)
