def on_click_inside(self, event, ax):
import plottool as pt
viztype = ph.get_plotdat(ax, 'viztype', None)
print('[ik] viztype=%r' % viztype)
if viztype is None:
pass
elif viztype == 'keypoints':
kpts = ph.get_plotdat(ax, 'kpts', [])
if len(kpts) == 0:
print('...nokpts')
else:
print('...nearest')
x, y = event.xdata, event.ydata
fx = ut.nearest_point(x, y, kpts)[0]
self._select_ith_kpt(fx)
elif viztype == 'warped':
hs_fx = ph.get_plotdat(ax, 'fx', None)
if hs_fx is not None:
kp = self.kpts[hs_fx] # FIXME
sift = self.vecs[hs_fx]
df2.draw_keypoint_gradient_orientations(self.chip,
kp,
sift=sift,
mode='vec',
fnum=pt.next_fnum())
pt.draw()
elif viztype.startswith('colorbar'):
pass
else:
print('...unhandled')
self.draw()
def on_click_inside(self, event, ax):
import plottool as pt
viztype = ph.get_plotdat(ax, 'viztype', None)
print('[ik] viztype=%r' % viztype)
if viztype is None:
pass
elif viztype == 'keypoints':
kpts = ph.get_plotdat(ax, 'kpts', [])
if len(kpts) == 0:
print('...nokpts')
else:
print('...nearest')
x, y = event.xdata, event.ydata
fx = ut.nearest_point(x, y, kpts)[0]
self._select_ith_kpt(fx)
elif viztype == 'warped':
hs_fx = ph.get_plotdat(ax, 'fx', None)
if hs_fx is not None:
kp = self.kpts[hs_fx] # FIXME
sift = self.vecs[hs_fx]
df2.draw_keypoint_gradient_orientations(self.chip, kp,
sift=sift, mode='vec',
fnum=pt.next_fnum())
pt.draw()
elif viztype.startswith('colorbar'):
pass
else:
print('...unhandled')
self.draw()
def show_keypoint_gradient_orientations(ibs, aid, fx, fnum=None, pnum=None, config2_=None):
# Draw the gradient vectors of a patch overlaying the keypoint
if fnum is None:
fnum = df2.next_fnum()
rchip = ibs.get_annot_chips(aid, config2_=config2_)
kp = ibs.get_annot_kpts(aid, config2_=config2_)[fx]
sift = ibs.get_annot_vecs(aid, config2_=config2_)[fx]
df2.draw_keypoint_gradient_orientations(rchip, kp, sift=sift,
mode='vec', fnum=fnum, pnum=pnum)
df2.set_title('Gradient orientation\n %s, fx=%d' % (get_aidstrs(aid), fx))
def _on_keypoints_click(event):
print('[viz] clicked keypoint view')
if event is None or event.xdata is None or event.inaxes is None:
annote_ptr[0] = (annote_ptr[0] + 1) % 3
mode = annote_ptr[0]
ell = mode == 1
pts = mode == 2
print('... default kpts view mode=%r' % mode)
_viz_keypoints(fnum, ell=ell, pts=pts,
**kwargs) # MAYBE: remove kwargs
else:
ax = event.inaxes
viztype = ph.get_plotdat(ax, 'viztype', None)
print('[ik] viztype=%r' % viztype)
if viztype == 'keypoints':
kpts = ph.get_plotdat(ax, 'kpts', [])
if len(kpts) == 0:
print('...nokpts')
else:
print('...nearest')
x, y = event.xdata, event.ydata
fx = ut.nearest_point(x, y, kpts)[0]
_select_ith_kpt(fx)
elif viztype == 'warped':
hs_fx = ph.get_plotdat(ax, 'fx', None)
#kpts = ph.get_plotdat(ax, 'kpts', [])
if hs_fx is not None:
# Ugly. Interactions should be changed to classes.
kp = self.kpts[hs_fx] # FIXME
sift = self.vecs[hs_fx]
df2.draw_keypoint_gradient_orientations(
chip, kp, sift=sift, mode='vec', fnum=df2.next_fnum())
elif viztype.startswith('colorbar'):
pass
# Hack to get a specific scoring feature
#sortx = self.fs.argsort()
#idx = np.clip(int(np.round(y * len(sortx))), 0, len(sortx) - 1)
#mx = sortx[idx]
#(fx1, fx2) = self.fm[mx]
#(fx1, fx2) = self.fm[mx]
#print('... selected score at rank idx=%r' % (idx,))
#print('... selected score with fs=%r' % (self.fs[mx],))
#print('... resolved to mx=%r' % mx)
#print('... fx1, fx2 = %r, %r' % (fx1, fx2,))
#self.select_ith_match(mx)
else:
print('...unhandled')
ph.draw()
def _on_keypoints_click(event):
print('[viz] clicked keypoint view')
if event is None or event.xdata is None or event.inaxes is None:
annote_ptr[0] = (annote_ptr[0] + 1) % 3
mode = annote_ptr[0]
ell = mode == 1
pts = mode == 2
print('... default kpts view mode=%r' % mode)
_viz_keypoints(fnum, ell=ell, pts=pts, **kwargs) # MAYBE: remove kwargs
else:
ax = event.inaxes
viztype = ph.get_plotdat(ax, 'viztype', None)
print('[ik] viztype=%r' % viztype)
if viztype == 'keypoints':
kpts = ph.get_plotdat(ax, 'kpts', [])
if len(kpts) == 0:
print('...nokpts')
else:
print('...nearest')
x, y = event.xdata, event.ydata
fx = ut.nearest_point(x, y, kpts)[0]
_select_ith_kpt(fx)
elif viztype == 'warped':
hs_fx = ph.get_plotdat(ax, 'fx', None)
#kpts = ph.get_plotdat(ax, 'kpts', [])
if hs_fx is not None:
# Ugly. Interactions should be changed to classes.
kp = self.kpts[hs_fx] # FIXME
sift = self.vecs[hs_fx]
df2.draw_keypoint_gradient_orientations(chip, kp, sift=sift, mode='vec',
fnum=df2.next_fnum())
elif viztype.startswith('colorbar'):
pass
# Hack to get a specific scoring feature
#sortx = self.fs.argsort()
#idx = np.clip(int(np.round(y * len(sortx))), 0, len(sortx) - 1)
#mx = sortx[idx]
#(fx1, fx2) = self.fm[mx]
#(fx1, fx2) = self.fm[mx]
#print('... selected score at rank idx=%r' % (idx,))
#print('... selected score with fs=%r' % (self.fs[mx],))
#print('... resolved to mx=%r' % mx)
#print('... fx1, fx2 = %r, %r' % (fx1, fx2,))
#self.select_ith_match(mx)
else:
print('...unhandled')
ph.draw()
def show_keypoint_gradient_orientations(ibs,
aid,
fx,
fnum=None,
pnum=None,
config2_=None):
# Draw the gradient vectors of a patch overlaying the keypoint
if fnum is None:
fnum = df2.next_fnum()
rchip = ibs.get_annot_chips(aid, config2_=config2_)
kp = ibs.get_annot_kpts(aid, config2_=config2_)[fx]
sift = ibs.get_annot_vecs(aid, config2_=config2_)[fx]
df2.draw_keypoint_gradient_orientations(rchip,
kp,
sift=sift,
mode='vec',
fnum=fnum,
pnum=pnum)
df2.set_title('Gradient orientation\n %s, fx=%d' % (get_aidstrs(aid), fx))
def draw_feat_row(
chip,
fx,
kp,
sift,
fnum,
nRows,
nCols=None,
px=None,
prevsift=None,
origsift=None,
aid=None,
info="",
type_=None,
shape_labels=False,
vecfield=False,
multicolored_arms=False,
draw_chip=False,
draw_warped=True,
draw_unwarped=True,
draw_desc=True,
rect=True,
ori=True,
pts=False,
**kwargs
):
"""
draw_feat_row
SeeAlso:
ibeis.viz.viz_nearest_descriptors
~/code/ibeis/ibeis/viz/viz_nearest_descriptors.py
CommandLine:
# Use this to find the fx you want to visualize
python -m plottool.interact_keypoints --test-ishow_keypoints --show --fname zebra.png
# Use this to visualize the featrow
python -m plottool.viz_featrow --test-draw_feat_row --show
python -m plottool.viz_featrow --test-draw_feat_row --show --fname zebra.png --fx=121 --feat-all --no-sift
python -m plottool.viz_featrow --test-draw_feat_row --dpath figures --save ~/latex/crall-candidacy-2015/figures/viz_featrow.jpg
Example:
>>> # DISABLE_DOCTEST
>>> from plottool.viz_featrow import * # NOQA
>>> import plottool as pt
>>> # build test data
>>> kpts, vecs, imgBGR = pt.viz_keypoints.testdata_kpts()
>>> chip = imgBGR
>>> print('There are %d features' % (len(vecs)))
>>> fx = ut.get_argval('--fx', type_=int, default=0)
>>> kp = kpts[fx]
>>> sift = vecs[fx]
>>> fnum = 1
>>> nRows = 1
>>> nCols = 2
>>> px = 0
>>> hack = ut.get_argflag('--feat-all')
>>> sift = sift if not ut.get_argflag('--no-sift') else None
>>> draw_desc = sift is not None
>>> kw = dict(
>>> prevsift=None, origsift=None, aid=None, info='', type_=None,
>>> shape_labels=False, vecfield=False, multicolored_arms=True,
>>> draw_chip=hack, draw_unwarped=hack, draw_warped=True, draw_desc=draw_desc
>>> )
>>> # execute function
>>> result = draw_feat_row(chip, fx, kp, sift, fnum, nRows, nCols, px,
>>> rect=False, ori=False, pts=False, **kw)
>>> # verify results
>>> print(result)
>>> pt.show_if_requested()
"""
import numpy as np
import vtool as vt
# should not need ncols here
if nCols is not None:
if ut.VERBOSE:
print("Warning nCols is no longer needed")
# assert nCols_ == nCols
nCols = draw_chip + draw_unwarped + draw_warped + draw_desc
pnum_ = df2.make_pnum_nextgen(nRows, nCols, start=px)
# pnum_ = df2.get_pnum_func(nRows, nCols, base=1)
# countgen = itertools.count(1)
# pnumgen_ = df2.make_pnum_nextgen(nRows, nCols, base=1)
def _draw_patch(**kwargs):
return df2.draw_keypoint_patch(
chip,
kp,
sift,
rect=rect,
ori=ori,
pts=pts,
ori_color=custom_constants.DEEP_PINK,
multicolored_arms=multicolored_arms,
**kwargs
)
# Feature strings
xy_str, shape_str, scale, ori_str = ph.kp_info(kp)
if draw_chip:
pnum = pnum_()
df2.imshow(chip, fnum=fnum, pnum=pnum)
kpts_kw = dict(ell_linewidth=5, ell_alpha=1.0)
kpts_kw.update(kwargs)
df2.draw_kpts2([kp], **kpts_kw)
if draw_unwarped:
# Draw the unwarped selected feature
# ax = _draw_patch(fnum=fnum, pnum=pnum_(px + six.next(countgen)))
# pnum = pnum_(px + six.next(countgen)
pnum = pnum_()
ax = _draw_patch(fnum=fnum, pnum=pnum)
ph.set_plotdat(ax, "viztype", "unwarped")
ph.set_plotdat(ax, "aid", aid)
ph.set_plotdat(ax, "fx", fx)
if shape_labels:
unwarped_lbl = "affine feature invV =\n" + shape_str + "\n" + ori_str
custom_figure.set_xlabel(unwarped_lbl, ax)
if draw_warped:
# Draw the warped selected feature
# ax = _draw_patch(fnum=fnum, pnum=pnum_(px + six.next(countgen)), warped=True)
pnum = pnum_()
ax = _draw_patch(fnum=fnum, pnum=pnum, warped=True, **kwargs)
ph.set_plotdat(ax, "viztype", "warped")
ph.set_plotdat(ax, "aid", aid)
ph.set_plotdat(ax, "fx", fx)
if shape_labels:
warped_lbl = ("warped feature\n" + "fx=%r scale=%.1f\n" + "%s") % (fx, scale, xy_str)
else:
warped_lbl = ""
warped_lbl += info
custom_figure.set_xlabel(warped_lbl, ax)
if draw_desc:
border_color = {
"None": None,
"query": None,
"match": custom_constants.BLUE,
"norm": custom_constants.ORANGE,
}.get(str(type_).lower(), None)
if border_color is not None:
df2.draw_border(ax, color=border_color)
# Draw the SIFT representation
# pnum = pnum_(px + six.next(countgen))
pnum = pnum_()
sift_as_vecfield = ph.SIFT_OR_VECFIELD or vecfield
if sift_as_vecfield:
custom_figure.figure(fnum=fnum, pnum=pnum)
df2.draw_keypoint_gradient_orientations(chip, kp, sift=sift)
else:
if sift.dtype.type == np.uint8:
sigtitle = "sift histogram" if (px % 3) == 0 else ""
ax = df2.plot_sift_signature(sift, sigtitle, fnum=fnum, pnum=pnum)
else:
sigtitle = "descriptor vector" if (px % 3) == 0 else ""
ax = df2.plot_descriptor_signature(sift, sigtitle, fnum=fnum, pnum=pnum)
ax._hs_viztype = "histogram"
# dist_list = ['L1', 'L2', 'hist_isect', 'emd']
# dist_list = ['L2', 'hist_isect']
# dist_list = ['L2']
# dist_list = ['bar_L2_sift', 'cos_sift']
# dist_list = ['L2_sift', 'bar_cos_sift']
dist_list = ["L2_sift"]
dist_str_list = []
if origsift is not None:
distmap_orig = vt.compute_distances(sift, origsift, dist_list)
dist_str_list.append(
"query_dist: "
+ ", ".join(["(%s, %s)" % (key, formatdist(val)) for key, val in six.iteritems(distmap_orig)])
)
if prevsift is not None:
distmap_prev = vt.compute_distances(sift, prevsift, dist_list)
dist_str_list.append(
"prev_dist: "
+ ", ".join(["(%s, %s)" % (key, formatdist(val)) for key, val in six.iteritems(distmap_prev)])
)
dist_str = "\n".join(dist_str_list)
custom_figure.set_xlabel(dist_str)
return px + nCols
def draw_feat_row(chip,
fx,
kp,
sift,
fnum,
nRows,
nCols=None,
px=None,
prevsift=None,
origsift=None,
aid=None,
info='',
type_=None,
shape_labels=False,
vecfield=False,
multicolored_arms=False,
draw_chip=False,
draw_warped=True,
draw_unwarped=True,
draw_desc=True,
rect=True,
ori=True,
pts=False,
**kwargs):
"""
draw_feat_row
SeeAlso:
ibeis.viz.viz_nearest_descriptors
~/code/ibeis/ibeis/viz/viz_nearest_descriptors.py
CommandLine:
# Use this to find the fx you want to visualize
python -m plottool.interact_keypoints --test-ishow_keypoints --show --fname zebra.png
# Use this to visualize the featrow
python -m plottool.viz_featrow --test-draw_feat_row --show
python -m plottool.viz_featrow --test-draw_feat_row --show --fname zebra.png --fx=121 --feat-all --no-sift
python -m plottool.viz_featrow --test-draw_feat_row --dpath figures --save ~/latex/crall-candidacy-2015/figures/viz_featrow.jpg
Example:
>>> # DISABLE_DOCTEST
>>> from plottool.viz_featrow import * # NOQA
>>> import plottool as pt
>>> # build test data
>>> kpts, vecs, imgBGR = pt.viz_keypoints.testdata_kpts()
>>> chip = imgBGR
>>> print('There are %d features' % (len(vecs)))
>>> fx = ut.get_argval('--fx', type_=int, default=0)
>>> kp = kpts[fx]
>>> sift = vecs[fx]
>>> fnum = 1
>>> nRows = 1
>>> nCols = 2
>>> px = 0
>>> hack = ut.get_argflag('--feat-all')
>>> sift = sift if not ut.get_argflag('--no-sift') else None
>>> draw_desc = sift is not None
>>> kw = dict(
>>> prevsift=None, origsift=None, aid=None, info='', type_=None,
>>> shape_labels=False, vecfield=False, multicolored_arms=True,
>>> draw_chip=hack, draw_unwarped=hack, draw_warped=True, draw_desc=draw_desc
>>> )
>>> # execute function
>>> result = draw_feat_row(chip, fx, kp, sift, fnum, nRows, nCols, px,
>>> rect=False, ori=False, pts=False, **kw)
>>> # verify results
>>> print(result)
>>> pt.show_if_requested()
"""
import numpy as np
import vtool as vt
# should not need ncols here
if nCols is not None:
if ut.VERBOSE:
print('Warning nCols is no longer needed')
#assert nCols_ == nCols
nCols = (draw_chip + draw_unwarped + draw_warped + draw_desc)
pnum_ = df2.make_pnum_nextgen(nRows, nCols, start=px)
#pnum_ = df2.get_pnum_func(nRows, nCols, base=1)
#countgen = itertools.count(1)
#pnumgen_ = df2.make_pnum_nextgen(nRows, nCols, base=1)
def _draw_patch(**kwargs):
return df2.draw_keypoint_patch(chip,
kp,
sift,
rect=rect,
ori=ori,
pts=pts,
ori_color=custom_constants.DEEP_PINK,
multicolored_arms=multicolored_arms,
**kwargs)
# Feature strings
xy_str, shape_str, scale, ori_str = ph.kp_info(kp)
if draw_chip:
pnum = pnum_()
df2.imshow(chip, fnum=fnum, pnum=pnum)
kpts_kw = dict(ell_linewidth=5, ell_alpha=1.0)
kpts_kw.update(kwargs)
df2.draw_kpts2([kp], **kpts_kw)
if draw_unwarped:
# Draw the unwarped selected feature
#ax = _draw_patch(fnum=fnum, pnum=pnum_(px + six.next(countgen)))
#pnum = pnum_(px + six.next(countgen)
pnum = pnum_()
ax = _draw_patch(fnum=fnum, pnum=pnum)
ph.set_plotdat(ax, 'viztype', 'unwarped')
ph.set_plotdat(ax, 'aid', aid)
ph.set_plotdat(ax, 'fx', fx)
if shape_labels:
unwarped_lbl = 'affine feature invV =\n' + shape_str + '\n' + ori_str
custom_figure.set_xlabel(unwarped_lbl, ax)
if draw_warped:
# Draw the warped selected feature
#ax = _draw_patch(fnum=fnum, pnum=pnum_(px + six.next(countgen)), warped=True)
pnum = pnum_()
ax = _draw_patch(fnum=fnum, pnum=pnum, warped=True, **kwargs)
ph.set_plotdat(ax, 'viztype', 'warped')
ph.set_plotdat(ax, 'aid', aid)
ph.set_plotdat(ax, 'fx', fx)
if shape_labels:
warped_lbl = ('warped feature\n' + 'fx=%r scale=%.1f\n' +
'%s') % (fx, scale, xy_str)
else:
warped_lbl = ''
warped_lbl += info
custom_figure.set_xlabel(warped_lbl, ax)
if draw_desc:
border_color = {
'None': None,
'query': None,
'match': custom_constants.BLUE,
'norm': custom_constants.ORANGE
}.get(str(type_).lower(), None)
if border_color is not None:
df2.draw_border(ax, color=border_color)
# Draw the SIFT representation
#pnum = pnum_(px + six.next(countgen))
pnum = pnum_()
sift_as_vecfield = ph.SIFT_OR_VECFIELD or vecfield
if sift_as_vecfield:
custom_figure.figure(fnum=fnum, pnum=pnum)
df2.draw_keypoint_gradient_orientations(chip, kp, sift=sift)
else:
if sift.dtype.type == np.uint8:
sigtitle = 'sift histogram' if (px % 3) == 0 else ''
ax = df2.plot_sift_signature(sift,
sigtitle,
fnum=fnum,
pnum=pnum)
else:
sigtitle = 'descriptor vector' if (px % 3) == 0 else ''
ax = df2.plot_descriptor_signature(sift,
sigtitle,
fnum=fnum,
pnum=pnum)
ax._hs_viztype = 'histogram'
#dist_list = ['L1', 'L2', 'hist_isect', 'emd']
#dist_list = ['L2', 'hist_isect']
#dist_list = ['L2']
#dist_list = ['bar_L2_sift', 'cos_sift']
#dist_list = ['L2_sift', 'bar_cos_sift']
dist_list = ['L2_sift']
dist_str_list = []
if origsift is not None:
distmap_orig = vt.compute_distances(sift, origsift, dist_list)
dist_str_list.append('query_dist: ' + ', '.join([
'(%s, %s)' % (key, formatdist(val))
for key, val in six.iteritems(distmap_orig)
]))
if prevsift is not None:
distmap_prev = vt.compute_distances(sift, prevsift, dist_list)
dist_str_list.append('prev_dist: ' + ', '.join([
'(%s, %s)' % (key, formatdist(val))
for key, val in six.iteritems(distmap_prev)
]))
dist_str = '\n'.join(dist_str_list)
custom_figure.set_xlabel(dist_str)
return px + nCols
