def show_nearest_descriptors(hs, qcx, qfx, fnum=None):
if fnum is None:
fnum = df2.next_fnum()
# Inspect the nearest neighbors of a descriptor
dx2_cx = hs.qdat._data_index.ax2_cx
dx2_fx = hs.qdat._data_index.ax2_fx
K = hs.qdat.cfg.nn_cfg.K
Knorm = hs.qdat.cfg.nn_cfg.Knorm
checks = hs.qdat.cfg.nn_cfg.checks
flann = hs.qdat._data_index.flann
qfx2_desc = hs.get_desc(qcx)[qfx:qfx + 1]
try:
(qfx2_dx, qfx2_dist) = flann.nn_index(qfx2_desc, K + Knorm, checks=checks)
qfx2_cx = dx2_cx[qfx2_dx]
qfx2_fx = dx2_fx[qfx2_dx]
def get_extract_tuple(cx, fx, k=-1):
rchip = hs.get_chip(cx)
kp = hs.get_kpts(cx)[fx]
sift = hs.get_desc(cx)[fx]
if k == -1:
info = '\nquery %s, fx=%r' % (hs.cidstr(cx), fx)
type_ = 'query'
elif k < K:
type_ = 'match'
info = '\nmatch %s, fx=%r k=%r, dist=%r' % (hs.cidstr(cx), fx, k, qfx2_dist[0, k])
elif k < Knorm + K:
type_ = 'norm'
info = '\nnorm %s, fx=%r k=%r, dist=%r' % (hs.cidstr(cx), fx, k, qfx2_dist[0, k])
else:
raise Exception('[viz] problem k=%r')
return (rchip, kp, sift, fx, cx, info, type_)
extracted_list = []
extracted_list.append(get_extract_tuple(qcx, qfx, -1))
for k in xrange(K + Knorm):
tup = get_extract_tuple(qfx2_cx[0, k], qfx2_fx[0, k], k)
extracted_list.append(tup)
#print('[viz] K + Knorm = %r' % (K + Knorm))
# Draw the _select_ith_match plot
nRows, nCols = len(extracted_list), 3
# Draw selected feature matches
prevsift = None
df2.figure(fnum=fnum, docla=True, doclf=True)
px = 0 # plot offset
for (rchip, kp, sift, fx, cx, info, type_) in extracted_list:
print('[viz] ' + info.replace('\n', ''))
px = draw_feat_row(rchip, fx, kp, sift, fnum, nRows, nCols, px,
prevsift=prevsift, cx=cx, info=info, type_=type_)
prevsift = sift
df2.adjust_subplots_safe(hspace=1)
except Exception as ex:
print('[viz] Error in show nearest descriptors')
print(ex)
raise
def plot_rank_histogram(allres, orgres_type):
print('[viz] plotting %r rank histogram' % orgres_type)
ranks = allres.__dict__[orgres_type].ranks
label = 'P(rank | ' + orgres_type + ' match)'
title = orgres_type + ' match rankings histogram\n' + allres.title_suffix
df2.figure(fnum=FIGNUM, doclf=True, title=title)
df2.draw_histpdf(ranks, label=label) # FIXME
df2.set_xlabel('ground truth ranks')
df2.set_ylabel('frequency')
df2.legend()
__dump_or_browse(allres.hs, 'rankviz')
def _chipmatch_view(pnum=(1, 1, 1), **kwargs):
mode = annote_ptr[0]
draw_ell = mode >= 1
draw_lines = mode == 2
annote_ptr[0] = (annote_ptr[0] + 1) % 3
df2.figure(fnum=fnum, docla=True, doclf=True)
# TODO RENAME This to remove res and rectify with show_chipres
tup = viz.res_show_chipres(res, hs, cx, fnum=fnum, pnum=pnum,
draw_lines=draw_lines, draw_ell=draw_ell,
colorbar_=True, **kwargs)
ax, xywh1, xywh2 = tup
xywh2_ptr[0] = xywh2
df2.set_figtitle(figtitle + hs.vs_str(qcx, cx))
def plot_score_pdf(allres, orgres_type, colorx=0.0, variation_truncate=False):
print('[viz] plotting ' + orgres_type + ' score pdf')
title = orgres_type + ' match score frequencies\n' + allres.title_suffix
scores = allres.__dict__[orgres_type].scores
print('[viz] len(scores) = %r ' % (len(scores), ))
label = 'P(score | %r)' % orgres_type
df2.figure(fnum=FIGNUM, doclf=True, title=title)
df2.draw_pdf(scores, label=label, colorx=colorx)
if variation_truncate:
df2.variation_trunctate(scores)
#df2.variation_trunctate(false.scores)
df2.set_xlabel('score')
df2.set_ylabel('frequency')
df2.legend()
__dump_or_browse(allres.hs, 'scoreviz')
def plot_rank_stem(allres, orgres_type='true'):
print('[viz] plotting rank stem')
# Visualize rankings with the stem plot
hs = allres.hs
title = orgres_type + 'rankings stem plot\n' + allres.title_suffix
orgres = allres.__dict__[orgres_type]
df2.figure(fnum=FIGNUM, doclf=True, title=title)
x_data = orgres.qcxs
y_data = orgres.ranks
df2.draw_stems(x_data, y_data)
slice_num = int(np.ceil(np.log10(len(orgres.qcxs))))
df2.set_xticks(hs.test_sample_cx[::slice_num])
df2.set_xlabel('query chip indeX (qcx)')
df2.set_ylabel('groundtruth chip ranks')
#df2.set_yticks(list(seen_ranks))
__dump_or_browse(allres.hs, 'rankviz')
def plot_name(hs, nx, nx2_cxs=None, fignum=0, **kwargs):
print('[viz*] plot_name nx=%r' % nx)
if not 'fignum' in vars():
kwargs = {}
fignum = 0
nx2_name = hs.tables.nx2_name
cx2_nx = hs.tables.cx2_nx
name = nx2_name[nx]
if not nx2_cxs is None:
cxs = nx2_cxs[nx]
else:
cxs = np.where(cx2_nx == nx)[0]
print('[viz*] plot_name %r' % hs.cxstr(cxs))
ncxs = len(cxs)
nCols = int(min(np.ceil(np.sqrt(ncxs)), 5))
nRows = int(np.ceil(ncxs / nCols))
print('[viz*] r=%r, c=%r' % (nRows, nCols))
gs2 = gridspec.GridSpec(nRows, nCols)
fig = df2.figure(fignum=fignum, **kwargs)
fig.clf()
for ss, cx in zip(gs2, cxs):
ax = fig.add_subplot(ss)
plot_cx(hs, cx)
title = 'nx=%r -- name=%r' % (nx, name)
#gs2.tight_layout(fig)
#gs2.update(top=df2.TOP_SUBPLOT_ADJUST)
df2.set_figtitle(title)
def show_name(hs, nx, nx2_cxs=None, fnum=0, sel_cxs=[], subtitle='',
annote=False, **kwargs):
print('[viz] show_name nx=%r' % nx)
nx2_name = hs.tables.nx2_name
cx2_nx = hs.tables.cx2_nx
name = nx2_name[nx]
if not nx2_cxs is None:
cxs = nx2_cxs[nx]
else:
cxs = np.where(cx2_nx == nx)[0]
print('[viz] show_name %r' % hs.cidstr(cxs))
nRows, nCols = get_square_row_cols(len(cxs))
print('[viz*] r=%r, c=%r' % (nRows, nCols))
#gs2 = gridspec.GridSpec(nRows, nCols)
pnum = lambda px: (nRows, nCols, px + 1)
fig = df2.figure(fnum=fnum, pnum=pnum(0), **kwargs)
fig.clf()
# Trigger computation of all chips in parallel
hs.refresh_features(cxs)
for px, cx in enumerate(cxs):
show_chip(hs, cx=cx, pnum=pnum(px), draw_ell=annote, kpts_alpha=.2)
if cx in sel_cxs:
ax = df2.gca()
df2.draw_border(ax, df2.GREEN, 4)
#plot_cx3(hs, cx)
if isinstance(nx, np.ndarray):
nx = nx[0]
if isinstance(name, np.ndarray):
name = name[0]
figtitle = 'Name View nx=%r name=%r' % (nx, name)
df2.set_figtitle(figtitle)
def plot_name(hs, nx, nx2_cxs=None, fignum=0, **kwargs):
print('[viz*] plot_name nx=%r' % nx)
if not 'fignum' in vars():
kwargs = {}
fignum = 0
nx2_name = hs.tables.nx2_name
cx2_nx = hs.tables.cx2_nx
name = nx2_name[nx]
if not nx2_cxs is None:
cxs = nx2_cxs[nx]
else:
cxs = np.where(cx2_nx == nx)[0]
print('[viz*] plot_name %r' % hs.cxstr(cxs))
ncxs = len(cxs)
nCols = int(min(np.ceil(np.sqrt(ncxs)), 5))
nRows = int(np.ceil(ncxs / nCols))
print('[viz*] r=%r, c=%r' % (nRows, nCols))
gs2 = gridspec.GridSpec(nRows, nCols)
fig = df2.figure(fignum=fignum, **kwargs)
fig.clf()
for ss, cx in zip(gs2, cxs):
ax = fig.add_subplot(ss)
plot_cx(hs, cx)
title = 'nx=%r -- name=%r' % (nx, name)
#gs2.tight_layout(fig)
#gs2.update(top=df2.TOP_SUBPLOT_ADJUST)
df2.set_figtitle(title)
def draw_splash(self):
print('[back] draw_splash()')
img = imread('_frontend/splash.png')
fig = df2.figure()
print(fig)
print(fig is self.win.plotWidget.figure)
df2.imshow(img)
df2.update()
def draw_splash(self):
print('[back] draw_splash()')
img = imread('_frontend/splash.png')
fig = df2.figure()
print(fig)
print(fig is self.win.plotWidget.figure)
df2.imshow(img)
df2.update()
def _ctrlclicked_cx(cx):
printDBG('ctrl+clicked cx=%r' % cx)
fnum = FNUMS['special']
fig = df2.figure(fnum=fnum, docla=True, doclf=True)
df2.disconnect_callback(fig, 'button_press_event')
viz_spatial_verification(hs, res.qcx, cx2=cx, fnum=fnum)
fig.canvas.draw()
df2.bring_to_front(fig)
def _ctrlclicked_cx(cx):
printDBG('ctrl+clicked cx=%r' % cx)
fnum = FNUMS['special']
fig = df2.figure(fnum=fnum, docla=True, doclf=True)
df2.disconnect_callback(fig, 'button_press_event')
viz_spatial_verification(hs, res.qcx, cx2=cx, fnum=fnum)
fig.canvas.draw()
df2.bring_to_front(fig)
def _select_ith_match(mx, qcx, cx):
#----------------------
# Get info for the _select_ith_match plot
annote_ptr[0] = 1
# Get the mx-th feature match
cx1, cx2 = qcx, cx
fx1, fx2 = fm[mx]
fscore2 = res.cx2_fs[cx2][mx]
fk2 = res.cx2_fk[cx2][mx]
kpts1, kpts2 = hs.get_kpts([cx1, cx2])
desc1, desc2 = hs.get_desc([cx1, cx2])
kp1, kp2 = kpts1[fx1], kpts2[fx2]
sift1, sift2 = desc1[fx1], desc2[fx2]
info1 = '\nquery'
info2 = '\nk=%r fscore=%r' % (fk2, fscore2)
last_state.last_fx = fx1
# Extracted keypoints to draw
extracted_list = [(rchip1, kp1, sift1, fx1, cx1, info1),
(rchip2, kp2, sift2, fx2, cx2, info2)]
# Normalizng Keypoint
if hasattr(res, 'filt2_meta') and 'lnbnn' in res.filt2_meta:
qfx2_norm = res.filt2_meta['lnbnn']
# Normalizing chip and feature
(cx3, fx3, normk) = qfx2_norm[fx1]
rchip3 = hs.get_chip(cx3)
kp3 = hs.get_kpts(cx3)[fx3]
sift3 = hs.get_desc(cx3)[fx3]
info3 = '\nnorm %s k=%r' % (hs.cidstr(cx3), normk)
extracted_list.append((rchip3, kp3, sift3, fx3, cx3, info3))
else:
print('WARNING: meta doesnt exist')
#----------------------
# Draw the _select_ith_match plot
nRows, nCols = len(extracted_list) + same_fig, 3
# Draw matching chips and features
sel_fm = np.array([(fx1, fx2)])
pnum1 = (nRows, 1, 1) if same_fig else (1, 1, 1)
_chipmatch_view(pnum1, vert=False, ell_alpha=.4, ell_linewidth=1.8,
colors=df2.BLUE, sel_fm=sel_fm, **kwargs)
# Draw selected feature matches
px = nCols * same_fig # plot offset
prevsift = None
if not same_fig:
fnum2 = fnum + len(viz.FNUMS)
fig2 = df2.figure(fnum=fnum2, docla=True, doclf=True)
else:
fnum2 = fnum
for (rchip, kp, sift, fx, cx, info) in extracted_list:
px = viz.draw_feat_row(rchip, fx, kp, sift, fnum2, nRows, nCols, px,
prevsift=prevsift, cx=cx, info=info)
prevsift = sift
if not same_fig:
df2.connect_callback(fig2, 'button_press_event', _click_chipres_click)
df2.set_figtitle(figtitle + hs.vs_str(qcx, cx))
def test3():
desc = np.random.rand(128)
desc = desc / np.sqrt((desc**2).sum())
desc = np.round(desc * 255)
print desc
fig = df2.figure(fignum=43)
ax = df2.plt.gca()
ax.set_xlim(-1,1)
ax.set_ylim(-1,1)
df2.draw_sift(desc)
def test(hs, qcx):
fx2_scale = sv2.keypoint_scale(hs.feats.cx2_kpts[qcx])
fx = fx2_scale.argsort()[::-1][40]
rchip = hs.get_chip(qcx)
kp = hs.feats.cx2_kpts[qcx][fx]
# Show full image and keypoint
df2.figure(fignum=9000, doclf=True)
df2.imshow(rchip, plotnum=(1,3,1))
df2.draw_kpts2([kp], ell_color=(1,0,0), pts=True)
# Show cropped image and keypoint
patch, subkp = get_patch(rchip, kp)
df2.imshow(patch, plotnum=(1,3,2))
df2.draw_kpts2([subkp], ell_color=(1,0,0), pts=True)
# Show warped image and keypoint
wpatch, wkp = get_warped_patch(rchip, kp)
df2.imshow(wpatch, plotnum=(1,3,3))
df2.draw_kpts2([wkp], ell_color=(1,0,0), pts=True)
#
df2.set_figtitle('warp test')
def dump_orgres_matches(allres, orgres_type):
orgres = allres.__dict__[orgres_type]
hs = allres.hs
qcx2_res = allres.qcx2_res
# loop over each query / result of interest
for qcx, cx, score, rank in orgres.iter():
query_gname, _ = os.path.splitext(
hs.tables.gx2_gname[hs.tables.cx2_gx[qcx]])
result_gname, _ = os.path.splitext(
hs.tables.gx2_gname[hs.tables.cx2_gx[cx]])
res = qcx2_res[qcx]
df2.figure(fnum=FIGNUM, pnum=121)
df2.show_matches3(res, hs, cx, SV=False, fnum=FIGNUM, pnum=121)
df2.show_matches3(res, hs, cx, SV=True, fnum=FIGNUM, pnum=122)
big_title = 'score=%.2f_rank=%d_q=%s_r=%s' % (score, rank, query_gname,
result_gname)
df2.set_figtitle(big_title)
__dump_or_browse(allres.hs,
orgres_type + '_matches' + allres.title_suffix)
def show_image(hs, gx, sel_cxs=[], fnum=1, figtitle='Img', annote=True,
draw_lbls=True, **kwargs):
# Shows an image with annotations
gname = hs.tables.gx2_gname[gx]
title = 'gx=%r gname=%r' % (gx, gname)
img = hs.gx2_image(gx)
fig = df2.figure(fnum=fnum, docla=True)
fig, ax = df2.imshow(img, title=title, fnum=fnum, **kwargs)
ax._hs_viewtype = 'image'
_annotate_image(hs, ax, gx, sel_cxs, draw_lbls, annote)
df2.set_figtitle(figtitle)
def plot_tt_bt_tf_matches(hs, allres, qcx):
#print('Visualizing result: ')
#res.printme()
res = allres.qcx2_res[qcx]
ranks = (allres.top_true_qcx_arrays[0][qcx],
allres.bot_true_qcx_arrays[0][qcx],
allres.top_false_qcx_arrays[0][qcx])
#scores = (allres.top_true_qcx_arrays[1][qcx],
#allres.bot_true_qcx_arrays[1][qcx],
#allres.top_false_qcx_arrays[1][qcx])
cxs = (allres.top_true_qcx_arrays[2][qcx],
allres.bot_true_qcx_arrays[2][qcx],
allres.top_false_qcx_arrays[2][qcx])
titles = ('best True rank=' + str(ranks[0]) + ' ',
'worst True rank=' + str(ranks[1]) + ' ',
'best False rank=' + str(ranks[2]) + ' ')
df2.figure(fnum=1, pnum=231)
res.plot_matches(res,
hs,
cxs[0],
False,
fnum=1,
pnum=131,
title_aug=titles[0])
res.plot_matches(res,
hs,
cxs[1],
False,
fnum=1,
pnum=132,
title_aug=titles[1])
res.plot_matches(res,
hs,
cxs[2],
False,
fnum=1,
pnum=133,
title_aug=titles[2])
fig_title = 'fig q' + hs.cidstr(
qcx) + ' TT BT TF -- ' + allres.title_suffix
df2.set_figtitle(fig_title)
def plot_score_matrix(allres):
print('[viz] plotting score matrix')
score_matrix = allres.score_matrix
title = 'Score Matrix\n' + allres.title_suffix
# Find inliers
#inliers = helpers.find_std_inliers(score_matrix)
#max_inlier = score_matrix[inliers].max()
# Trunate above 255
score_img = np.copy(score_matrix)
#score_img[score_img < 0] = 0
#score_img[score_img > 255] = 255
#dim = 0
#score_img = helpers.norm_zero_one(score_img, dim=dim)
# Make colors
scores = score_img.flatten()
colors = df2.scores_to_color(scores, logscale=True)
cmap = df2.scores_to_cmap(scores, colors)
df2.figure(fnum=FIGNUM, doclf=True, title=title)
# Show score matrix
df2.imshow(score_img, fnum=FIGNUM, cmap=cmap)
# Colorbar
df2.colorbar(scores, colors)
df2.set_xlabel('database')
df2.set_ylabel('queries')
def show_image(hs,
gx,
sel_cxs=[],
fnum=1,
figtitle='Img',
annote=True,
draw_lbls=True,
**kwargs):
# Shows an image with annotations
gname = hs.tables.gx2_gname[gx]
title = 'gx=%r gname=%r' % (gx, gname)
img = hs.gx2_image(gx)
fig = df2.figure(fnum=fnum, docla=True)
fig, ax = df2.imshow(img, title=title, fnum=fnum, **kwargs)
ax._hs_viewtype = 'image'
_annotate_image(hs, ax, gx, sel_cxs, draw_lbls, annote)
df2.set_figtitle(figtitle)
def show_name(hs,
nx,
nx2_cxs=None,
fnum=0,
sel_cxs=[],
subtitle='',
annote=False,
**kwargs):
print('[viz] show_name nx=%r' % nx)
nx2_name = hs.tables.nx2_name
cx2_nx = hs.tables.cx2_nx
name = nx2_name[nx]
if not nx2_cxs is None:
cxs = nx2_cxs[nx]
else:
cxs = np.where(cx2_nx == nx)[0]
print('[viz] show_name %r' % hs.cidstr(cxs))
nRows, nCols = get_square_row_cols(len(cxs))
print('[viz*] r=%r, c=%r' % (nRows, nCols))
#gs2 = gridspec.GridSpec(nRows, nCols)
pnum = lambda px: (nRows, nCols, px + 1)
fig = df2.figure(fnum=fnum, pnum=pnum(0), **kwargs)
fig.clf()
# Trigger computation of all chips in parallel
hs.refresh_features(cxs)
for px, cx in enumerate(cxs):
show_chip(hs, cx=cx, pnum=pnum(px), draw_ell=annote, kpts_alpha=.2)
if cx in sel_cxs:
ax = df2.gca()
df2.draw_border(ax, df2.GREEN, 4)
#plot_cx3(hs, cx)
if isinstance(nx, np.ndarray):
nx = nx[0]
if isinstance(name, np.ndarray):
name = name[0]
figtitle = 'Name View nx=%r name=%r' % (nx, name)
df2.set_figtitle(figtitle)
def viz_spatial_verification(hs,
cx1,
figtitle='Spatial Verification View',
**kwargs):
#kwargs = {}
from hscom import helpers
from hotspotter import spatial_verification2 as sv2
import cv2
print('\n======================')
cx2 = ensure_cx2(hs, cx1, kwargs.pop('cx2', None))
print('[viz] viz_spatial_verification %s' % hs.vs_str(cx1, cx2))
fnum = kwargs.get('fnum', 4)
fm = ensure_fm(hs, cx1, cx2, kwargs.pop('fm', None),
kwargs.pop('res', 'db'))
# Get keypoints
rchip1 = kwargs['rchip1'] if 'rchip1' in kwargs else hs.get_chip(cx1)
rchip2 = kwargs['rchip2'] if 'rchip1' in kwargs else hs.get_chip(cx2)
kpts1 = kwargs['kpts1'] if 'kpts1' in kwargs else hs.get_kpts(cx1)
kpts2 = kwargs['kpts2'] if 'kpts2' in kwargs else hs.get_kpts(cx2)
dlen_sqrd2 = rchip2.shape[0]**2 + rchip2.shape[1]**2
# rchips are in shape = (height, width)
(h1, w1) = rchip1.shape[0:2]
(h2, w2) = rchip2.shape[0:2]
#wh1 = (w1, h1)
wh2 = (w2, h2)
#print('[viz.sv] wh1 = %r' % (wh1,))
#print('[viz.sv] wh2 = %r' % (wh2,))
# Get affine and homog mapping from rchip1 to rchip2
xy_thresh = hs.prefs.query_cfg.sv_cfg.xy_thresh
max_scale = hs.prefs.query_cfg.sv_cfg.scale_thresh_high
min_scale = hs.prefs.query_cfg.sv_cfg.scale_thresh_low
homog_args = [
kpts1, kpts2, fm, xy_thresh, max_scale, min_scale, dlen_sqrd2, 4
]
try:
Aff, aff_inliers = sv2.homography_inliers(*homog_args,
just_affine=True)
H, inliers = sv2.homography_inliers(*homog_args, just_affine=False)
except Exception as ex:
print('[viz] homog_args = %r' % (homog_args))
print('[viz] ex = %r' % (ex, ))
raise
print(helpers.horiz_string(['H = ', str(H)]))
print(helpers.horiz_string(['Aff = ', str(Aff)]))
# Transform the chips
print('warp homog')
rchip1_Ht = cv2.warpPerspective(rchip1, H, wh2)
print('warp affine')
rchip1_At = cv2.warpAffine(rchip1, Aff[0:2, :], wh2)
rchip2_blendA = np.zeros(rchip2.shape, dtype=rchip2.dtype)
rchip2_blendH = np.zeros(rchip2.shape, dtype=rchip2.dtype)
rchip2_blendA = rchip2 / 2 + rchip1_At / 2
rchip2_blendH = rchip2 / 2 + rchip1_Ht / 2
df2.figure(fnum=fnum, pnum=(3, 4, 1), docla=True, doclf=True)
def _draw_chip(title, chip, px, *args, **kwargs):
df2.imshow(chip,
*args,
title=title,
fnum=fnum,
pnum=(3, 4, px),
**kwargs)
# Draw original matches, affine inliers, and homography inliers
def _draw_matches(title, fm, px):
# Helper with common arguments to df2.show_chipmatch2
dmkwargs = dict(fs=None,
title=title,
all_kpts=False,
draw_lines=True,
docla=True,
fnum=fnum,
pnum=(3, 3, px))
df2.show_chipmatch2(rchip1,
rchip2,
kpts1,
kpts2,
fm,
show_nMatches=True,
**dmkwargs)
# Draw the Assigned -> Affine -> Homography matches
_draw_matches('Assigned matches', fm, 1)
_draw_matches('Affine inliers', fm[aff_inliers], 2)
_draw_matches('Homography inliers', fm[inliers], 3)
# Draw the Affine Transformations
_draw_chip('Source', rchip1, 5)
_draw_chip('Affine', rchip1_At, 6)
_draw_chip('Destination', rchip2, 7)
_draw_chip('Aff Blend', rchip2_blendA, 8)
# Draw the Homography Transformation
_draw_chip('Source', rchip1, 9)
_draw_chip('Homog', rchip1_Ht, 10)
_draw_chip('Destination', rchip2, 11)
_draw_chip('Homog Blend', rchip2_blendH, 12)
df2.set_figtitle(figtitle)
def _show_res(hs, res, **kwargs):
''' Displays query chip, groundtruth matches, and top 5 matches'''
#printDBG('[viz._show_res()] %s ' % helpers.printableVal(locals()))
#printDBG = print
in_image = hs.prefs.display_cfg.show_results_in_image
annote = kwargs.pop('annote', 2) # this is toggled
fnum = kwargs.get('fnum', 3)
figtitle = kwargs.get('figtitle', '')
topN_cxs = kwargs.get('topN_cxs', [])
gt_cxs = kwargs.get('gt_cxs', [])
all_kpts = kwargs.get('all_kpts', False)
interact = kwargs.get('interact', True)
show_query = kwargs.get('show_query', False)
dosquare = kwargs.get('dosquare', False)
if SHOW_QUERY_OVERRIDE is not None:
show_query = SHOW_QUERY_OVERRIDE
max_nCols = 5
if len(topN_cxs) in [6, 7]:
max_nCols = 3
if len(topN_cxs) in [8]:
max_nCols = 4
printDBG('========================')
printDBG('[viz._show_res()]----------------')
all_gts = hs.get_other_indexed_cxs(res.qcx)
_tup = tuple(map(len, (topN_cxs, gt_cxs, all_gts)))
printDBG('[viz._show_res()] #topN=%r #missed_gts=%r/%r' % _tup)
printDBG('[viz._show_res()] * fnum=%r' % (fnum,))
printDBG('[viz._show_res()] * figtitle=%r' % (figtitle,))
printDBG('[viz._show_res()] * max_nCols=%r' % (max_nCols,))
printDBG('[viz._show_res()] * show_query=%r' % (show_query,))
ranked_cxs = res.topN_cxs(hs, N='all')
# Build a subplot grid
nQuerySubplts = 1 if show_query else 0
nGtSubplts = nQuerySubplts + (0 if gt_cxs is None else len(gt_cxs))
nTopNSubplts = 0 if topN_cxs is None else len(topN_cxs)
nTopNCols = min(max_nCols, nTopNSubplts)
nGTCols = min(max_nCols, nGtSubplts)
nGTCols = max(nGTCols, nTopNCols)
nTopNCols = nGTCols
nGtRows = 0 if nGTCols == 0 else int(np.ceil(nGtSubplts / nGTCols))
nTopNRows = 0 if nTopNCols == 0 else int(np.ceil(nTopNSubplts / nTopNCols))
nGtCells = nGtRows * nGTCols
nRows = nTopNRows + nGtRows
# HACK:
_color_list = df2.distinct_colors(len(topN_cxs))
cx2_color = {cx: _color_list[ox] for ox, cx in enumerate(topN_cxs)}
if IN_IMAGE_OVERRIDE is not None:
in_image = IN_IMAGE_OVERRIDE
# Helpers
def _show_query_fn(plotx_shift, rowcols):
'helper for viz._show_res'
plotx = plotx_shift + 1
pnum = (rowcols[0], rowcols[1], plotx)
#print('[viz] Plotting Query: pnum=%r' % (pnum,))
_kwshow = dict(draw_kpts=annote)
_kwshow.update(kwargs)
_kwshow['prefix'] = 'q'
_kwshow['res'] = res
_kwshow['pnum'] = pnum
_kwshow['cx2_color'] = cx2_color
_kwshow['draw_ell'] = annote >= 1
#_kwshow['in_image'] = in_image
show_chip(hs, **_kwshow)
#if in_image:
#roi1 = hs.cx2_roi(res.qcx)
#df2.draw_roi(roi1, bbox_color=df2.ORANGE, label='q' + hs.cidstr(res.qcx))
def _plot_matches_cxs(cx_list, plotx_shift, rowcols):
def _show_matches_fn(cx, orank, pnum):
'Helper function for drawing matches to one cx'
aug = 'rank=%r\n' % orank
#printDBG('[viz._show_res()] plotting: %r' % (pnum,))
_kwshow = dict(draw_ell=annote, draw_pts=False, draw_lines=annote,
ell_alpha=.5, all_kpts=all_kpts)
_kwshow.update(kwargs)
_kwshow['fnum'] = fnum
_kwshow['pnum'] = pnum
_kwshow['title_aug'] = aug
# If we already are showing the query dont show it here
if not show_query:
_kwshow['draw_ell'] = annote == 1
_kwshow['draw_lines'] = annote >= 1
res_show_chipres(res, hs, cx, in_image=in_image, **_kwshow)
else:
_kwshow['draw_ell'] = annote >= 1
if annote == 2:
# TODO Find a better name
_kwshow['color'] = cx2_color[cx]
_kwshow['sel_fx2'] = res.cx2_fm[cx][:, 1]
show_chip(hs, cx, in_image=in_image, **_kwshow)
annotate_chipres(hs, res, cx, in_image=in_image, show_query=not show_query)
'helper for viz._show_res to draw many cxs'
#printDBG('[viz._show_res()] Plotting Chips %s:' % hs.cidstr(cx_list))
if cx_list is None:
return
for ox, cx in enumerate(cx_list):
plotx = ox + plotx_shift + 1
pnum = (rowcols[0], rowcols[1], plotx)
oranks = np.where(ranked_cxs == cx)[0]
if len(oranks) == 0:
orank = -1
continue
orank = oranks[0] + 1
_show_matches_fn(cx, orank, pnum)
if dosquare:
# HACK
nSubplots = nGtSubplts + nTopNSubplts
nRows, nCols = get_square_row_cols(nSubplots, 3)
nTopNCols = nGTCols = nCols
shift_topN = 1
printDBG('nRows, nCols = (%r, %r)' % (nRows, nCols))
else:
shift_topN = nGtCells
if nGtSubplts == 1:
nGTCols = 1
fig = df2.figure(fnum=fnum, pnum=(nRows, nGTCols, 1), docla=True, doclf=True)
df2.disconnect_callback(fig, 'button_press_event')
df2.plt.subplot(nRows, nGTCols, 1)
# Plot Query
if show_query:
_show_query_fn(0, (nRows, nGTCols))
# Plot Ground Truth
_plot_matches_cxs(gt_cxs, nQuerySubplts, (nRows, nGTCols))
_plot_matches_cxs(topN_cxs, shift_topN, (nRows, nTopNCols))
figtitle += ' q%s name=%s' % (hs.cidstr(res.qcx), hs.cx2_name(res.qcx))
df2.set_figtitle(figtitle, incanvas=not NO_LABEL_OVERRIDE)
# Result Interaction
if interact:
printDBG('[viz._show_res()] starting interaction')
def _ctrlclicked_cx(cx):
printDBG('ctrl+clicked cx=%r' % cx)
fnum = FNUMS['special']
fig = df2.figure(fnum=fnum, docla=True, doclf=True)
df2.disconnect_callback(fig, 'button_press_event')
viz_spatial_verification(hs, res.qcx, cx2=cx, fnum=fnum)
fig.canvas.draw()
df2.bring_to_front(fig)
def _clicked_cx(cx):
printDBG('clicked cx=%r' % cx)
fnum = FNUMS['inspect']
res.interact_chipres(hs, cx, fnum=fnum)
fig = df2.gcf()
fig.canvas.draw()
df2.bring_to_front(fig)
def _clicked_none():
# Toggle if the click is not in any axis
printDBG('clicked none')
#print(kwargs)
_show_res(hs, res, annote=(annote + 1) % 3, **kwargs)
fig.canvas.draw()
def _on_res_click(event):
'result interaction mpl event callback slot'
print('[viz] clicked result')
if event.xdata is None or event.inaxes is None:
#print('clicked outside axes')
return _clicked_none()
ax = event.inaxes
hs_viewtype = ax.__dict__.get('_hs_viewtype', '')
printDBG(event.__dict__)
printDBG('hs_viewtype=%r' % hs_viewtype)
# Clicked a specific chipres
if hs_viewtype.find('chipres') == 0:
cx = ax.__dict__.get('_hs_cx')
# Ctrl-Click
key = '' if event.key is None else event.key
print('key = %r' % key)
if key.find('control') == 0:
print('[viz] result control clicked')
return _ctrlclicked_cx(cx)
# Left-Click
else:
print('[viz] result clicked')
return _clicked_cx(cx)
df2.connect_callback(fig, 'button_press_event', _on_res_click)
df2.adjust_subplots_safe()
printDBG('[viz._show_res()] Finished')
return fig
ax.add_collection(ellipse_collection) rchip_fpath = 'tpl/extern_feat/lena.png' rchip_fpath = 'tpl/extern_feat/zebra.jpg' rchip_fpath = os.path.realpath(rchip_fpath) rchip = cv2.cvtColor(cv2.imread(rchip_fpath, flags=cv2.IMREAD_COLOR), cv2.COLOR_BGR2RGB) outname = extern_feat.compute_perdoch_text_feats(rchip_fpath) #outname = compute_inria_text_feats(rchip_fpath, 'harris', 'sift') # Keep the wrong way to compare kpts0, desc = extern_feat.read_text_feat_file(outname) invE = extern_feat.expand_invET(kpts0[:,2:5].T)[0] kpts1 = extern_feat.fix_kpts_hack(kpts0[:], method=1) A1 = extern_feat.expand_acd(kpts1[:,2:5])[0] #kpts, desc = filter_kpts_scale(kpts, desc) df2.figure(1, doclf=True) df2.DARKEN = .5 #---- df2.imshow(rchip, plotnum=(2,3,1), title='0 before (inverse square root)') draw_kpts3(kpts0.copy(),0) #---- df2.imshow(rchip, plotnum=(2,3,2), title='1 just inverse') draw_kpts3(kpts1.copy(),1) #---- df2.imshow(rchip, plotnum=(2,3,3), title='2 identity') draw_kpts3(kpts1.copy(), 2) df2.update() #----
def show_nearest_descriptors(hs, qcx, qfx, fnum=None):
if fnum is None:
fnum = df2.next_fnum()
# Inspect the nearest neighbors of a descriptor
dx2_cx = hs.qdat._data_index.ax2_cx
dx2_fx = hs.qdat._data_index.ax2_fx
K = hs.qdat.cfg.nn_cfg.K
Knorm = hs.qdat.cfg.nn_cfg.Knorm
checks = hs.qdat.cfg.nn_cfg.checks
flann = hs.qdat._data_index.flann
qfx2_desc = hs.get_desc(qcx)[qfx:qfx + 1]
try:
(qfx2_dx, qfx2_dist) = flann.nn_index(qfx2_desc,
K + Knorm,
checks=checks)
qfx2_cx = dx2_cx[qfx2_dx]
qfx2_fx = dx2_fx[qfx2_dx]
def get_extract_tuple(cx, fx, k=-1):
rchip = hs.get_chip(cx)
kp = hs.get_kpts(cx)[fx]
sift = hs.get_desc(cx)[fx]
if k == -1:
info = '\nquery %s, fx=%r' % (hs.cidstr(cx), fx)
type_ = 'query'
elif k < K:
type_ = 'match'
info = '\nmatch %s, fx=%r k=%r, dist=%r' % (hs.cidstr(cx), fx,
k, qfx2_dist[0, k])
elif k < Knorm + K:
type_ = 'norm'
info = '\nnorm %s, fx=%r k=%r, dist=%r' % (hs.cidstr(cx), fx,
k, qfx2_dist[0, k])
else:
raise Exception('[viz] problem k=%r')
return (rchip, kp, sift, fx, cx, info, type_)
extracted_list = []
extracted_list.append(get_extract_tuple(qcx, qfx, -1))
for k in xrange(K + Knorm):
tup = get_extract_tuple(qfx2_cx[0, k], qfx2_fx[0, k], k)
extracted_list.append(tup)
#print('[viz] K + Knorm = %r' % (K + Knorm))
# Draw the _select_ith_match plot
nRows, nCols = len(extracted_list), 3
# Draw selected feature matches
prevsift = None
df2.figure(fnum=fnum, docla=True, doclf=True)
px = 0 # plot offset
for (rchip, kp, sift, fx, cx, info, type_) in extracted_list:
print('[viz] ' + info.replace('\n', ''))
px = draw_feat_row(rchip,
fx,
kp,
sift,
fnum,
nRows,
nCols,
px,
prevsift=prevsift,
cx=cx,
info=info,
type_=type_)
prevsift = sift
df2.adjust_subplots_safe(hspace=1)
except Exception as ex:
print('[viz] Error in show nearest descriptors')
print(ex)
raise
def viz_spatial_verification(hs, cx1, figtitle='Spatial Verification View', **kwargs):
#kwargs = {}
from hscom import helpers
from hotspotter import spatial_verification2 as sv2
import cv2
print('\n======================')
cx2 = ensure_cx2(hs, cx1, kwargs.pop('cx2', None))
print('[viz] viz_spatial_verification %s' % hs.vs_str(cx1, cx2))
fnum = kwargs.get('fnum', 4)
fm = ensure_fm(hs, cx1, cx2, kwargs.pop('fm', None), kwargs.pop('res', 'db'))
# Get keypoints
rchip1 = kwargs['rchip1'] if 'rchip1' in kwargs else hs.get_chip(cx1)
rchip2 = kwargs['rchip2'] if 'rchip1' in kwargs else hs.get_chip(cx2)
kpts1 = kwargs['kpts1'] if 'kpts1' in kwargs else hs.get_kpts(cx1)
kpts2 = kwargs['kpts2'] if 'kpts2' in kwargs else hs.get_kpts(cx2)
dlen_sqrd2 = rchip2.shape[0] ** 2 + rchip2.shape[1] ** 2
# rchips are in shape = (height, width)
(h1, w1) = rchip1.shape[0:2]
(h2, w2) = rchip2.shape[0:2]
#wh1 = (w1, h1)
wh2 = (w2, h2)
#print('[viz.sv] wh1 = %r' % (wh1,))
#print('[viz.sv] wh2 = %r' % (wh2,))
# Get affine and homog mapping from rchip1 to rchip2
xy_thresh = hs.prefs.query_cfg.sv_cfg.xy_thresh
max_scale = hs.prefs.query_cfg.sv_cfg.scale_thresh_high
min_scale = hs.prefs.query_cfg.sv_cfg.scale_thresh_low
homog_args = [kpts1, kpts2, fm, xy_thresh, max_scale, min_scale, dlen_sqrd2, 4]
try:
Aff, aff_inliers = sv2.homography_inliers(*homog_args, just_affine=True)
H, inliers = sv2.homography_inliers(*homog_args, just_affine=False)
except Exception as ex:
print('[viz] homog_args = %r' % (homog_args))
print('[viz] ex = %r' % (ex,))
raise
print(helpers.horiz_string(['H = ', str(H)]))
print(helpers.horiz_string(['Aff = ', str(Aff)]))
# Transform the chips
print('warp homog')
rchip1_Ht = cv2.warpPerspective(rchip1, H, wh2)
print('warp affine')
rchip1_At = cv2.warpAffine(rchip1, Aff[0:2, :], wh2)
rchip2_blendA = np.zeros(rchip2.shape, dtype=rchip2.dtype)
rchip2_blendH = np.zeros(rchip2.shape, dtype=rchip2.dtype)
rchip2_blendA = rchip2 / 2 + rchip1_At / 2
rchip2_blendH = rchip2 / 2 + rchip1_Ht / 2
df2.figure(fnum=fnum, pnum=(3, 4, 1), docla=True, doclf=True)
def _draw_chip(title, chip, px, *args, **kwargs):
df2.imshow(chip, *args, title=title, fnum=fnum, pnum=(3, 4, px), **kwargs)
# Draw original matches, affine inliers, and homography inliers
def _draw_matches(title, fm, px):
# Helper with common arguments to df2.show_chipmatch2
dmkwargs = dict(fs=None, title=title, all_kpts=False, draw_lines=True,
docla=True, fnum=fnum, pnum=(3, 3, px))
df2.show_chipmatch2(rchip1, rchip2, kpts1, kpts2, fm, show_nMatches=True, **dmkwargs)
# Draw the Assigned -> Affine -> Homography matches
_draw_matches('Assigned matches', fm, 1)
_draw_matches('Affine inliers', fm[aff_inliers], 2)
_draw_matches('Homography inliers', fm[inliers], 3)
# Draw the Affine Transformations
_draw_chip('Source', rchip1, 5)
_draw_chip('Affine', rchip1_At, 6)
_draw_chip('Destination', rchip2, 7)
_draw_chip('Aff Blend', rchip2_blendA, 8)
# Draw the Homography Transformation
_draw_chip('Source', rchip1, 9)
_draw_chip('Homog', rchip1_Ht, 10)
_draw_chip('Destination', rchip2, 11)
_draw_chip('Homog Blend', rchip2_blendH, 12)
df2.set_figtitle(figtitle)
def in_depth_ellipse2x2(rchip, kp):
#-----------------------
# SETUP
#-----------------------
from hotspotter import draw_func2 as df2
np.set_printoptions(precision=8)
tau = 2 * np.pi
df2.reset()
df2.figure(9003, docla=True, doclf=True)
ax = df2.gca()
ax.invert_yaxis()
def _plotpts(data, px, color=df2.BLUE, label=''):
#df2.figure(9003, docla=True, pnum=(1, 1, px))
df2.plot2(data.T[0], data.T[1], '.', '', color=color, label=label)
df2.update()
def _plotarrow(x, y, dx, dy, color=df2.BLUE, label=''):
ax = df2.gca()
arrowargs = dict(head_width=.5, length_includes_head=True, label=label)
arrow = df2.FancyArrow(x, y, dx, dy, **arrowargs)
arrow.set_edgecolor(color)
arrow.set_facecolor(color)
ax.add_patch(arrow)
df2.update()
def _2x2_eig(M2x2):
(evals, evecs) = np.linalg.eig(M2x2)
l1, l2 = evals
v1, v2 = evecs
return l1, l2, v1, v2
#-----------------------
# INPUT
#-----------------------
# We will call perdoch's invA = invV
print('--------------------------------')
print('Let V = Perdoch.A')
print('Let Z = Perdoch.E')
print('--------------------------------')
print('Input from Perdoch\'s detector: ')
# We are given the keypoint in invA format
(x, y, ia11, ia21, ia22), ia12 = kp, 0
invV = np.array([[ia11, ia12],
[ia21, ia22]])
V = np.linalg.inv(invV)
# <HACK>
#invV = V / np.linalg.det(V)
#V = np.linalg.inv(V)
# </HACK>
Z = (V.T).dot(V)
print('invV is a transform from points on a unit-circle to the ellipse')
helpers.horiz_print('invV = ', invV)
print('--------------------------------')
print('V is a transformation from points on the ellipse to a unit circle')
helpers.horiz_print('V = ', V)
print('--------------------------------')
print('Points on a matrix satisfy (x).T.dot(Z).dot(x) = 1')
print('where Z = (V.T).dot(V)')
helpers.horiz_print('Z = ', Z)
# Define points on a unit circle
theta_list = np.linspace(0, tau, 50)
cicrle_pts = np.array([(np.cos(t), np.sin(t)) for t in theta_list])
# Transform those points to the ellipse using invV
ellipse_pts1 = invV.dot(cicrle_pts.T).T
# Transform those points to the ellipse using V
ellipse_pts2 = V.dot(cicrle_pts.T).T
#Lets check our assertion: (x_).T.dot(Z).dot(x_) = 1
checks1 = [x_.T.dot(Z).dot(x_) for x_ in ellipse_pts1]
checks2 = [x_.T.dot(Z).dot(x_) for x_ in ellipse_pts2]
assert all([abs(1 - check) < 1E-11 for check in checks1])
#assert all([abs(1 - check) < 1E-11 for check in checks2])
print('... all of our plotted points satisfy this')
#=======================
# THE CONIC SECTION
#=======================
# All of this was from the Perdoch paper, now lets move into conic sections
# We will use the notation from wikipedia
# http://en.wikipedia.org/wiki/Conic_section
# http://en.wikipedia.org/wiki/Matrix_representation_of_conic_sections
#-----------------------
# MATRIX REPRESENTATION
#-----------------------
# The matrix representation of a conic is:
(A, B2, B2_, C) = Z.flatten()
(D, E, F) = (0, 0, 1)
B = B2 * 2
assert B2 == B2_, 'matrix should by symmetric'
print('--------------------------------')
print('Now, using wikipedia\' matrix representation of a conic.')
con = np.array(((' A', 'B / 2', 'D / 2'),
('B / 2', ' C', 'E / 2'),
('D / 2', 'E / 2', ' F')))
helpers.horiz_print('A matrix A_Q = ', con)
# A_Q is our conic section (aka ellipse matrix)
A_Q = np.array((( A, B / 2, D / 2),
(B / 2, C, E / 2),
(D / 2, E / 2, F)))
helpers.horiz_print('A_Q = ', A_Q)
#-----------------------
# DEGENERATE CONICS
#-----------------------
print('----------------------------------')
print('As long as det(A_Q) != it is not degenerate.')
print('If the conic is not degenerate, we can use the 2x2 minor: A_33')
print('det(A_Q) = %s' % str(np.linalg.det(A_Q)))
assert np.linalg.det(A_Q) != 0, 'degenerate conic'
A_33 = np.array((( A, B / 2),
(B / 2, C)))
helpers.horiz_print('A_33 = ', A_33)
#-----------------------
# CONIC CLASSIFICATION
#-----------------------
print('----------------------------------')
print('The determinant of the minor classifies the type of conic it is')
print('(det == 0): parabola, (det < 0): hyperbola, (det > 0): ellipse')
print('det(A_33) = %s' % str(np.linalg.det(A_33)))
assert np.linalg.det(A_33) > 0, 'conic is not an ellipse'
print('... this is indeed an ellipse')
#-----------------------
# CONIC CENTER
#-----------------------
print('----------------------------------')
print('the centers of the ellipse are obtained by: ')
print('x_center = (B * E - (2 * C * D)) / (4 * A * C - B ** 2)')
print('y_center = (D * B - (2 * A * E)) / (4 * A * C - B ** 2)')
# Centers are obtained by solving for where the gradient of the quadratic
# becomes 0. Without going through the derivation the calculation is...
# These should be 0, 0 if we are at the origin, or our original x, y
# coordinate specified by the keypoints. I'm doing the calculation just for
# shits and giggles
x_center = (B * E - (2 * C * D)) / (4 * A * C - B ** 2)
y_center = (D * B - (2 * A * E)) / (4 * A * C - B ** 2)
helpers.horiz_print('x_center = ', x_center)
helpers.horiz_print('y_center = ', y_center)
#-----------------------
# MAJOR AND MINOR AXES
#-----------------------
# Now we are going to determine the major and minor axis
# of this beast. It just the center augmented by the eigenvecs
print('----------------------------------')
# The angle between the major axis and our x axis is:
l1, l2, v1, v2 = _2x2_eig(A_33)
x_axis = np.array([1, 0])
theta = np.arccos(x_axis.dot(v1))
# The eccentricity is determined by:
nu = 1
numer = 2 * np.sqrt((A - C) ** 2 + B ** 2)
denom = nu * (A + C) + np.sqrt((A - C) ** 2 + B ** 2)
eccentricity = np.sqrt(numer / denom)
from scipy.special import ellipeinc
#-----------------------
# DRAWING
#-----------------------
# Lets start off by drawing the ellipse that we are goign to work with
# Create unit circle sample
# Draw the keypoint using the tried and true df2
# Other things should subsiquently align
df2.draw_kpts2(np.array([(0, 0, ia11, ia21, ia22)]), ell_linewidth=4,
ell_color=df2.DEEP_PINK, ell_alpha=1, arrow=True, rect=True)
# Plot ellipse points
_plotpts(ellipse_pts1, 0, df2.YELLOW, label='invV.dot(cicrle_pts.T).T')
# Plot ellipse axis
# !HELP! I DO NOT KNOW WHY I HAVE TO DIVIDE, SQUARE ROOT, AND NEGATE!!!
l1, l2, v1, v2 = _2x2_eig(A_33)
dx1, dy1 = (v1 / np.sqrt(l1))
dx2, dy2 = (v2 / np.sqrt(l2))
_plotarrow(0, 0, dx1, -dy1, color=df2.ORANGE, label='ellipse axis')
_plotarrow(0, 0, dx2, -dy2, color=df2.ORANGE)
# Plot ellipse orientation
orient_axis = invV.dot(np.eye(2))
dx1, dx2, dy1, dy2 = orient_axis.flatten()
_plotarrow(0, 0, dx1, dy1, color=df2.BLUE, label='ellipse rotation')
_plotarrow(0, 0, dx2, dy2, color=df2.BLUE)
df2.legend()
df2.dark_background()
df2.gca().invert_yaxis()
return locals()
def _plot(data, px, title=''):
df2.figure(9003, docla=True, pnum=(2, 4, px))
df2.plot2(data.T[0], data.T[1], '.', title)
def viz_top_features(hs, res, low, high, fignum=0, draw_chips=True):
from collections import defaultdict
qcx = res.qcx
cx2_nx = hs.tables.cx2_nx
top_patches_list = get_top_scoring_patches(hs, res, low, high)
num_rows = high-low
num_cols = 4
if params.__MATCH_TYPE__ == 'vsmany':
num_cols = 6
# Initialize Figure
fig = df2.figure(fignum+1, plotnum=(num_rows, num_cols,1))
cx2_rchip = defaultdict(int)
cx2_kplist = defaultdict(list)
def draw_one_kp(patch, kp, plotx, color, desc=None):
fig, ax = df2.imshow(patch, plotnum=(num_rows, num_cols, plotx))
df2.draw_kpts2([kp], ell_color=color, pts=True)
if not desc is None and not '--nodesc' in sys.argv:
df2.draw_sift(desc, [kp])
df2.draw_border(ax, color, 1)
for tx, (patches1, patches2, patchesN, cx, feat_score) in enumerate(top_patches_list):
(kp1, subkp1, wkp1, patch1, wpatch1, cx1, desc1) = patches1
(kp2, subkp2, wkp2, patch2, wpatch2, cx2, desc2) = patches2
# draw on table
# Draw Query Keypoint
plotx = (tx*num_cols)
draw_one_kp(patch1, subkp1, plotx+1, df2.GREEN, desc1)
qnx = cx2_nx[qcx]
df2.plt.gca().set_xlabel('qcx=%r; qnx=%r' % (qcx, qnx))
draw_one_kp(wpatch1, wkp1, plotx+2, df2.GREEN, desc1)
# Draw ith < k match
draw_one_kp(patch2, subkp2, plotx+3, df2.BLUE, desc2)
nx = cx2_nx[cx]
df2.plt.gca().set_xlabel('cx=%r; nx=%r' % (cx, nx))
draw_one_kp(wpatch2, wkp2, plotx+4, df2.BLUE, desc2)
df2.plt.gca().set_xlabel('score=%r' % (feat_score))
# Draw k+1th match
if params.__MATCH_TYPE__ == 'vsmany':
(kpN, subkpN, wkpN, patchN, wpatchN, cxN, descN) = patchesN
draw_one_kp(patchN, subkpN, plotx+5, df2.ORANGE, descN)
nxN = cx2_nx[qcx]
df2.plt.gca().set_xlabel('cxN=%r; nxN=%r' % (cxN, nxN))
draw_one_kp(wpatchN, wkpN, plotx+6, df2.ORANGE, descN)
# Get other info
cx2_rchip[cx] = hs.get_chip(cx)
cx2_kplist[qcx].append(kp1)
cx2_kplist[cx].append(kp2)
# Draw annotations on
df2.figure(plotnum=(num_rows, num_cols,1), title='Query Patch')
df2.figure(plotnum=(num_rows, num_cols,2), title='Query Patch')
df2.figure(plotnum=(num_rows, num_cols,3), title='Result Patch')
df2.figure(plotnum=(num_rows, num_cols,4), title='Result Warped')
if params.__MATCH_TYPE__ == 'vsmany':
df2.figure(plotnum=(num_rows, num_cols,5), title='Normalizer Patch')
df2.figure(plotnum=(num_rows, num_cols,6), title='Normalizer Warped')
df2.set_figtitle('Top '+str(low)+' to '+str(high)+' scoring matches')
if not draw_chips:
return
#
# Draw on full images
cx2_rchip[qcx] = hs.get_chip(qcx)
cx_keys = cx2_kplist.keys()
cx_vals = map(len, cx2_kplist.values())
cx_list = [x for (y,x) in sorted(zip(cx_vals, cx_keys))][::-1]
num_chips = len(cx_list)
pltnum_fn = lambda ix: (int(np.ceil(num_chips/2)), 2, ix)
plotnum = pltnum_fn(1)
fig2 = df2.figure(fignum-1, plotnum=plotnum)
for ix, cx in enumerate(cx_list):
plotnum = pltnum_fn(ix+1)
fig2 = df2.figure(fignum-1, plotnum=plotnum)
rchip = cx2_rchip[cx]
fig, ax = df2.imshow(rchip)
title_pref = ''
color = df2.BLUE
if res.qcx == cx:
title_pref = 'q'
color = df2.GREEN
df2.draw_kpts2(cx2_kplist[cx], ell_color=color, ell_alpha=1)
df2.draw_border(ax, color, 2)
nx = cx2_nx[cx]
ax.set_title(title_pref+'cx = %r; nx=%r' % (cx, nx))
gname = hs.get_gname(cx)
ax.set_xlabel(gname)
def _plot(data, px, title=''):
df2.figure(9003, docla=True, pnum=(2, 4, px))
df2.plot2(data.T[0], data.T[1], '.', title)
def _viz_keypoints(fnum, pnum=(1, 1, 1), **kwargs):
df2.figure(fnum=fnum, docla=True, doclf=True)
viz.show_keypoints(rchip, kpts, fnum=fnum, pnum=pnum, **kwargs)
if figtitle is not None:
df2.set_figtitle(figtitle)
def _show_res(hs, res, **kwargs):
''' Displays query chip, groundtruth matches, and top 5 matches'''
#printDBG('[viz._show_res()] %s ' % helpers.printableVal(locals()))
#printDBG = print
in_image = hs.prefs.display_cfg.show_results_in_image
annote = kwargs.pop('annote', 2) # this is toggled
fnum = kwargs.get('fnum', 3)
figtitle = kwargs.get('figtitle', '')
topN_cxs = kwargs.get('topN_cxs', [])
gt_cxs = kwargs.get('gt_cxs', [])
all_kpts = kwargs.get('all_kpts', False)
interact = kwargs.get('interact', True)
show_query = kwargs.get('show_query', False)
dosquare = kwargs.get('dosquare', False)
if SHOW_QUERY_OVERRIDE is not None:
show_query = SHOW_QUERY_OVERRIDE
max_nCols = 5
if len(topN_cxs) in [6, 7]:
max_nCols = 3
if len(topN_cxs) in [8]:
max_nCols = 4
printDBG('========================')
printDBG('[viz._show_res()]----------------')
all_gts = hs.get_other_indexed_cxs(res.qcx)
_tup = tuple(map(len, (topN_cxs, gt_cxs, all_gts)))
printDBG('[viz._show_res()] #topN=%r #missed_gts=%r/%r' % _tup)
printDBG('[viz._show_res()] * fnum=%r' % (fnum, ))
printDBG('[viz._show_res()] * figtitle=%r' % (figtitle, ))
printDBG('[viz._show_res()] * max_nCols=%r' % (max_nCols, ))
printDBG('[viz._show_res()] * show_query=%r' % (show_query, ))
ranked_cxs = res.topN_cxs(hs, N='all')
# Build a subplot grid
nQuerySubplts = 1 if show_query else 0
nGtSubplts = nQuerySubplts + (0 if gt_cxs is None else len(gt_cxs))
nTopNSubplts = 0 if topN_cxs is None else len(topN_cxs)
nTopNCols = min(max_nCols, nTopNSubplts)
nGTCols = min(max_nCols, nGtSubplts)
nGTCols = max(nGTCols, nTopNCols)
nTopNCols = nGTCols
nGtRows = 0 if nGTCols == 0 else int(np.ceil(nGtSubplts / nGTCols))
nTopNRows = 0 if nTopNCols == 0 else int(np.ceil(nTopNSubplts / nTopNCols))
nGtCells = nGtRows * nGTCols
nRows = nTopNRows + nGtRows
# HACK:
_color_list = df2.distinct_colors(len(topN_cxs))
cx2_color = {cx: _color_list[ox] for ox, cx in enumerate(topN_cxs)}
if IN_IMAGE_OVERRIDE is not None:
in_image = IN_IMAGE_OVERRIDE
# Helpers
def _show_query_fn(plotx_shift, rowcols):
'helper for viz._show_res'
plotx = plotx_shift + 1
pnum = (rowcols[0], rowcols[1], plotx)
#print('[viz] Plotting Query: pnum=%r' % (pnum,))
_kwshow = dict(draw_kpts=annote)
_kwshow.update(kwargs)
_kwshow['prefix'] = 'q'
_kwshow['res'] = res
_kwshow['pnum'] = pnum
_kwshow['cx2_color'] = cx2_color
_kwshow['draw_ell'] = annote >= 1
#_kwshow['in_image'] = in_image
show_chip(hs, **_kwshow)
#if in_image:
#roi1 = hs.cx2_roi(res.qcx)
#df2.draw_roi(roi1, bbox_color=df2.ORANGE, label='q' + hs.cidstr(res.qcx))
def _plot_matches_cxs(cx_list, plotx_shift, rowcols):
def _show_matches_fn(cx, orank, pnum):
'Helper function for drawing matches to one cx'
aug = 'rank=%r\n' % orank
#printDBG('[viz._show_res()] plotting: %r' % (pnum,))
_kwshow = dict(draw_ell=annote,
draw_pts=False,
draw_lines=annote,
ell_alpha=.5,
all_kpts=all_kpts)
_kwshow.update(kwargs)
_kwshow['fnum'] = fnum
_kwshow['pnum'] = pnum
_kwshow['title_aug'] = aug
# If we already are showing the query dont show it here
if not show_query:
_kwshow['draw_ell'] = annote == 1
_kwshow['draw_lines'] = annote >= 1
res_show_chipres(res, hs, cx, in_image=in_image, **_kwshow)
else:
_kwshow['draw_ell'] = annote >= 1
if annote == 2:
# TODO Find a better name
_kwshow['color'] = cx2_color[cx]
_kwshow['sel_fx2'] = res.cx2_fm[cx][:, 1]
show_chip(hs, cx, in_image=in_image, **_kwshow)
annotate_chipres(hs,
res,
cx,
in_image=in_image,
show_query=not show_query)
'helper for viz._show_res to draw many cxs'
#printDBG('[viz._show_res()] Plotting Chips %s:' % hs.cidstr(cx_list))
if cx_list is None:
return
for ox, cx in enumerate(cx_list):
plotx = ox + plotx_shift + 1
pnum = (rowcols[0], rowcols[1], plotx)
oranks = np.where(ranked_cxs == cx)[0]
if len(oranks) == 0:
orank = -1
continue
orank = oranks[0] + 1
_show_matches_fn(cx, orank, pnum)
if dosquare:
# HACK
nSubplots = nGtSubplts + nTopNSubplts
nRows, nCols = get_square_row_cols(nSubplots, 3)
nTopNCols = nGTCols = nCols
shift_topN = 1
printDBG('nRows, nCols = (%r, %r)' % (nRows, nCols))
else:
shift_topN = nGtCells
if nGtSubplts == 1:
nGTCols = 1
fig = df2.figure(fnum=fnum,
pnum=(nRows, nGTCols, 1),
docla=True,
doclf=True)
df2.disconnect_callback(fig, 'button_press_event')
df2.plt.subplot(nRows, nGTCols, 1)
# Plot Query
if show_query:
_show_query_fn(0, (nRows, nGTCols))
# Plot Ground Truth
_plot_matches_cxs(gt_cxs, nQuerySubplts, (nRows, nGTCols))
_plot_matches_cxs(topN_cxs, shift_topN, (nRows, nTopNCols))
figtitle += ' q%s name=%s' % (hs.cidstr(res.qcx), hs.cx2_name(res.qcx))
df2.set_figtitle(figtitle, incanvas=not NO_LABEL_OVERRIDE)
# Result Interaction
if interact:
printDBG('[viz._show_res()] starting interaction')
def _ctrlclicked_cx(cx):
printDBG('ctrl+clicked cx=%r' % cx)
fnum = FNUMS['special']
fig = df2.figure(fnum=fnum, docla=True, doclf=True)
df2.disconnect_callback(fig, 'button_press_event')
viz_spatial_verification(hs, res.qcx, cx2=cx, fnum=fnum)
fig.canvas.draw()
df2.bring_to_front(fig)
def _clicked_cx(cx):
printDBG('clicked cx=%r' % cx)
fnum = FNUMS['inspect']
res.interact_chipres(hs, cx, fnum=fnum)
fig = df2.gcf()
fig.canvas.draw()
df2.bring_to_front(fig)
def _clicked_none():
# Toggle if the click is not in any axis
printDBG('clicked none')
#print(kwargs)
_show_res(hs, res, annote=(annote + 1) % 3, **kwargs)
fig.canvas.draw()
def _on_res_click(event):
'result interaction mpl event callback slot'
print('[viz] clicked result')
if event.xdata is None or event.inaxes is None:
#print('clicked outside axes')
return _clicked_none()
ax = event.inaxes
hs_viewtype = ax.__dict__.get('_hs_viewtype', '')
printDBG(event.__dict__)
printDBG('hs_viewtype=%r' % hs_viewtype)
# Clicked a specific chipres
if hs_viewtype.find('chipres') == 0:
cx = ax.__dict__.get('_hs_cx')
# Ctrl-Click
key = '' if event.key is None else event.key
print('key = %r' % key)
if key.find('control') == 0:
print('[viz] result control clicked')
return _ctrlclicked_cx(cx)
# Left-Click
else:
print('[viz] result clicked')
return _clicked_cx(cx)
df2.connect_callback(fig, 'button_press_event', _on_res_click)
df2.adjust_subplots_safe()
printDBG('[viz._show_res()] Finished')
return fig
def in_depth_ellipse2x2(rchip, kp):
#-----------------------
# SETUP
#-----------------------
from hotspotter import draw_func2 as df2
np.set_printoptions(precision=8)
tau = 2 * np.pi
df2.reset()
df2.figure(9003, docla=True, doclf=True)
ax = df2.gca()
ax.invert_yaxis()
def _plotpts(data, px, color=df2.BLUE, label=''):
#df2.figure(9003, docla=True, pnum=(1, 1, px))
df2.plot2(data.T[0], data.T[1], '.', '', color=color, label=label)
df2.update()
def _plotarrow(x, y, dx, dy, color=df2.BLUE, label=''):
ax = df2.gca()
arrowargs = dict(head_width=.5, length_includes_head=True, label=label)
arrow = df2.FancyArrow(x, y, dx, dy, **arrowargs)
arrow.set_edgecolor(color)
arrow.set_facecolor(color)
ax.add_patch(arrow)
df2.update()
def _2x2_eig(M2x2):
(evals, evecs) = np.linalg.eig(M2x2)
l1, l2 = evals
v1, v2 = evecs
return l1, l2, v1, v2
#-----------------------
# INPUT
#-----------------------
# We will call perdoch's invA = invV
print('--------------------------------')
print('Let V = Perdoch.A')
print('Let Z = Perdoch.E')
print('--------------------------------')
print('Input from Perdoch\'s detector: ')
# We are given the keypoint in invA format
(x, y, ia11, ia21, ia22), ia12 = kp, 0
invV = np.array([[ia11, ia12], [ia21, ia22]])
V = np.linalg.inv(invV)
# <HACK>
#invV = V / np.linalg.det(V)
#V = np.linalg.inv(V)
# </HACK>
Z = (V.T).dot(V)
print('invV is a transform from points on a unit-circle to the ellipse')
helpers.horiz_print('invV = ', invV)
print('--------------------------------')
print('V is a transformation from points on the ellipse to a unit circle')
helpers.horiz_print('V = ', V)
print('--------------------------------')
print('Points on a matrix satisfy (x).T.dot(Z).dot(x) = 1')
print('where Z = (V.T).dot(V)')
helpers.horiz_print('Z = ', Z)
# Define points on a unit circle
theta_list = np.linspace(0, tau, 50)
cicrle_pts = np.array([(np.cos(t), np.sin(t)) for t in theta_list])
# Transform those points to the ellipse using invV
ellipse_pts1 = invV.dot(cicrle_pts.T).T
# Transform those points to the ellipse using V
ellipse_pts2 = V.dot(cicrle_pts.T).T
#Lets check our assertion: (x_).T.dot(Z).dot(x_) = 1
checks1 = [x_.T.dot(Z).dot(x_) for x_ in ellipse_pts1]
checks2 = [x_.T.dot(Z).dot(x_) for x_ in ellipse_pts2]
assert all([abs(1 - check) < 1E-11 for check in checks1])
#assert all([abs(1 - check) < 1E-11 for check in checks2])
print('... all of our plotted points satisfy this')
#=======================
# THE CONIC SECTION
#=======================
# All of this was from the Perdoch paper, now lets move into conic sections
# We will use the notation from wikipedia
# http://en.wikipedia.org/wiki/Conic_section
# http://en.wikipedia.org/wiki/Matrix_representation_of_conic_sections
#-----------------------
# MATRIX REPRESENTATION
#-----------------------
# The matrix representation of a conic is:
(A, B2, B2_, C) = Z.flatten()
(D, E, F) = (0, 0, 1)
B = B2 * 2
assert B2 == B2_, 'matrix should by symmetric'
print('--------------------------------')
print('Now, using wikipedia\' matrix representation of a conic.')
con = np.array(((' A', 'B / 2', 'D / 2'), ('B / 2', ' C', 'E / 2'),
('D / 2', 'E / 2', ' F')))
helpers.horiz_print('A matrix A_Q = ', con)
# A_Q is our conic section (aka ellipse matrix)
A_Q = np.array(((A, B / 2, D / 2), (B / 2, C, E / 2), (D / 2, E / 2, F)))
helpers.horiz_print('A_Q = ', A_Q)
#-----------------------
# DEGENERATE CONICS
#-----------------------
print('----------------------------------')
print('As long as det(A_Q) != it is not degenerate.')
print('If the conic is not degenerate, we can use the 2x2 minor: A_33')
print('det(A_Q) = %s' % str(np.linalg.det(A_Q)))
assert np.linalg.det(A_Q) != 0, 'degenerate conic'
A_33 = np.array(((A, B / 2), (B / 2, C)))
helpers.horiz_print('A_33 = ', A_33)
#-----------------------
# CONIC CLASSIFICATION
#-----------------------
print('----------------------------------')
print('The determinant of the minor classifies the type of conic it is')
print('(det == 0): parabola, (det < 0): hyperbola, (det > 0): ellipse')
print('det(A_33) = %s' % str(np.linalg.det(A_33)))
assert np.linalg.det(A_33) > 0, 'conic is not an ellipse'
print('... this is indeed an ellipse')
#-----------------------
# CONIC CENTER
#-----------------------
print('----------------------------------')
print('the centers of the ellipse are obtained by: ')
print('x_center = (B * E - (2 * C * D)) / (4 * A * C - B ** 2)')
print('y_center = (D * B - (2 * A * E)) / (4 * A * C - B ** 2)')
# Centers are obtained by solving for where the gradient of the quadratic
# becomes 0. Without going through the derivation the calculation is...
# These should be 0, 0 if we are at the origin, or our original x, y
# coordinate specified by the keypoints. I'm doing the calculation just for
# shits and giggles
x_center = (B * E - (2 * C * D)) / (4 * A * C - B**2)
y_center = (D * B - (2 * A * E)) / (4 * A * C - B**2)
helpers.horiz_print('x_center = ', x_center)
helpers.horiz_print('y_center = ', y_center)
#-----------------------
# MAJOR AND MINOR AXES
#-----------------------
# Now we are going to determine the major and minor axis
# of this beast. It just the center augmented by the eigenvecs
print('----------------------------------')
# The angle between the major axis and our x axis is:
l1, l2, v1, v2 = _2x2_eig(A_33)
x_axis = np.array([1, 0])
theta = np.arccos(x_axis.dot(v1))
# The eccentricity is determined by:
nu = 1
numer = 2 * np.sqrt((A - C)**2 + B**2)
denom = nu * (A + C) + np.sqrt((A - C)**2 + B**2)
eccentricity = np.sqrt(numer / denom)
from scipy.special import ellipeinc
#-----------------------
# DRAWING
#-----------------------
# Lets start off by drawing the ellipse that we are goign to work with
# Create unit circle sample
# Draw the keypoint using the tried and true df2
# Other things should subsiquently align
df2.draw_kpts2(np.array([(0, 0, ia11, ia21, ia22)]),
ell_linewidth=4,
ell_color=df2.DEEP_PINK,
ell_alpha=1,
arrow=True,
rect=True)
# Plot ellipse points
_plotpts(ellipse_pts1, 0, df2.YELLOW, label='invV.dot(cicrle_pts.T).T')
# Plot ellipse axis
# !HELP! I DO NOT KNOW WHY I HAVE TO DIVIDE, SQUARE ROOT, AND NEGATE!!!
l1, l2, v1, v2 = _2x2_eig(A_33)
dx1, dy1 = (v1 / np.sqrt(l1))
dx2, dy2 = (v2 / np.sqrt(l2))
_plotarrow(0, 0, dx1, -dy1, color=df2.ORANGE, label='ellipse axis')
_plotarrow(0, 0, dx2, -dy2, color=df2.ORANGE)
# Plot ellipse orientation
orient_axis = invV.dot(np.eye(2))
dx1, dx2, dy1, dy2 = orient_axis.flatten()
_plotarrow(0, 0, dx1, dy1, color=df2.BLUE, label='ellipse rotation')
_plotarrow(0, 0, dx2, dy2, color=df2.BLUE)
df2.legend()
df2.dark_background()
df2.gca().invert_yaxis()
return locals()
def get_kp_border(rchip, kp):
np.set_printoptions(precision=8)
df2.reset()
df2.figure(9003, docla=True, doclf=True)
def _plotpts(data, px, color=df2.BLUE, label=''):
#df2.figure(9003, docla=True, pnum=(1, 1, px))
df2.plot2(data.T[0], data.T[1], '-', '', color=color, label=label)
df2.update()
def _plotarrow(x, y, dx, dy, color=df2.BLUE, label=''):
ax = df2.gca()
arrowargs = dict(head_width=.5, length_includes_head=True, label='')
arrow = df2.FancyArrow(x, y, dx, dy, **arrowargs)
arrow.set_edgecolor(color)
arrow.set_facecolor(color)
ax.add_patch(arrow)
df2.update()
def _2x2_eig(M2x2):
(evals, evecs) = np.linalg.eig(M2x2)
l1, l2 = evals
v1, v2 = evecs
return l1, l2, v1, v2
#-----------------------
# INPUT
#-----------------------
# We are given the keypoint in invA format
(x, y, ia11, ia21, ia22), ia12 = kp, 0
# invA2x2 is a transformation from points on a unit circle to the ellipse
invA2x2 = np.array([[ia11, ia12], [ia21, ia22]])
#-----------------------
# DRAWING
#-----------------------
# Lets start off by drawing the ellipse that we are goign to work with
# Create unit circle sample
tau = 2 * np.pi
theta_list = np.linspace(0, tau, 1000)
cicrle_pts = np.array([(np.cos(t), np.sin(t)) for t in theta_list])
ellipse_pts = invA2x2.dot(cicrle_pts.T).T
_plotpts(ellipse_pts, 0, df2.BLACK, label='invA2x2.dot(unit_circle)')
l1, l2, v1, v2 = _2x2_eig(invA2x2)
dx1, dy1 = (v1 * l1)
dx2, dy2 = (v2 * l2)
_plotarrow(0, 0, dx1, dy1, color=df2.ORANGE, label='invA2x2 e1')
_plotarrow(0, 0, dx2, dy2, color=df2.RED, label='invA2x2 e2')
#-----------------------
# REPRESENTATION
#-----------------------
# A2x2 is a transformation from points on the ellipse to a unit circle
A2x2 = np.linalg.inv(invA2x2)
# Points on a matrix satisfy (x).T.dot(E2x2).dot(x) = 1
E2x2 = A2x2.T.dot(A2x2)
#Lets check our assertion: (x).T.dot(E2x2).dot(x) = 1
checks = [pt.T.dot(E2x2).dot(pt) for pt in ellipse_pts]
assert all([abs(1 - check) < 1E-11 for check in checks])
#-----------------------
# CONIC SECTIONS
#-----------------------
# All of this was from the Perdoch paper, now lets move into conic sections
# We will use the notation from wikipedia
# http://en.wikipedia.org/wiki/Conic_section
# http://en.wikipedia.org/wiki/Matrix_representation_of_conic_sections
# The matrix representation of a conic is:
((A, B, B_, C), (D, E, F)) = (E2x2.flatten(), (0, 0, 1))
assert B == B_, 'matrix should by symmetric'
# A_Q is our conic section (aka ellipse matrix)
A_Q = np.array(((A, B / 2, D / 2), (B / 2, C, E / 2), (D / 2, E / 2, F)))
assert np.linalg.det(A_Q) != 0, 'degenerate conic'
# As long as det(A_Q) is not 0 it is not degenerate and we can work with the
# minor 2x2 matrix
A_33 = np.array(((A, B / 2), (B / 2, C)))
# (det == 0)->parabola, (det < 0)->hyperbola, (det > 0)->ellipse
assert np.linalg.det(A_33) > 0, 'conic is not an ellipse'
# Centers are obtained by solving for where the gradient of the quadratic
# becomes 0. Without going through the derivation the calculation is...
# These should be 0, 0 if we are at the origin, or our original x, y
# coordinate specified by the keypoints. I'm doing the calculation just for
# shits and giggles
x_center = (B * E - (2 * C * D)) / (4 * A * C - B**2)
y_center = (D * B - (2 * A * E)) / (4 * A * C - B**2)
#=================
# DRAWING
#=================
# Now we are going to determine the major and minor axis
# of this beast. It just the center augmented by the eigenvecs
l1, l2, v1, v2 = _2x2_eig(A_33)
dx1, dy1 = 0 - (v1 / np.sqrt(l1))
dx2, dy2 = 0 - (v2 / np.sqrt(l2))
_plotarrow(0, 0, dx1, dy1, color=df2.BLUE)
_plotarrow(0, 0, dx2, dy2, color=df2.BLUE)
# The angle between the major axis and our x axis is:
x_axis = np.array([1, 0])
theta = np.arccos(x_axis.dot(evec1))
# The eccentricity is determined by:
nu = 1
numer = 2 * np.sqrt((A - C)**2 + B**2)
denom = nu * (A + C) + np.sqrt((A - C)**2 + B**2)
eccentricity = np.sqrt(numer / denom)
from scipy.special import ellipeinc
# Algebraic form of connic
#assert (a * (x ** 2)) + (b * (x * y)) + (c * (y ** 2)) + (d * x) + (e * y) + (f) == 0
#---------------------
invA = np.array([[a, 0], [c, d]])
Ashape = np.linalg.inv(np.array([[a, 0], [c, d]]))
Ashape /= np.sqrt(np.linalg.det(Ashape))
tau = 2 * np.pi
nSamples = 100
theta_list = np.linspace(0, tau, nSamples)
# Create unit circle sample
cicrle_pts = np.array([(np.cos(t), np.sin(t)) for t in theta_list])
circle_hpts = np.hstack([cicrle_pts, np.ones((len(cicrle_pts), 1))])
# Transform as if the unit cirle was the warped patch
ashape_pts = Ashape.dot(cicrle_pts.T).T
inv = np.linalg.inv
svd = np.linalg.svd
U, S_, V = svd(Ashape)
S = np.diag(S_)
pxl_list3 = invA.dot(cicrle_pts[:, 0:2].T).T
pxl_list4 = invA.dot(ashape_pts[:, 0:2].T).T
pxl_list5 = invA.T.dot(cicrle_pts[:, 0:2].T).T
pxl_list6 = invA.T.dot(ashape_pts[:, 0:2].T).T
pxl_list7 = inv(V).dot(ashape_pts[:, 0:2].T).T
pxl_list8 = inv(U).dot(ashape_pts[:, 0:2].T).T
df2.draw()
def _plot(data, px, title=''):
df2.figure(9003, docla=True, pnum=(2, 4, px))
df2.plot2(data.T[0], data.T[1], '.', title)
df2.figure(9003, doclf=True)
_plot(cicrle_pts, 1, 'unit circle')
_plot(ashape_pts, 2, 'A => circle shape')
_plot(pxl_list3, 3)
_plot(pxl_list4, 4)
_plot(pxl_list5, 5)
_plot(pxl_list6, 6)
_plot(pxl_list7, 7)
_plot(pxl_list8, 8)
df2.draw()
invA = np.array([[a, 0, x], [c, d, y], [0, 0, 1]])
pxl_list = invA.dot(circle_hpts.T).T[:, 0:2]
df2.figure(9002, doclf=True)
df2.imshow(rchip)
df2.plot2(pxl_list.T[0], pxl_list.T[1], '.')
df2.draw()
vals = [cv2.getRectSubPix(rchip, (1, 1), tuple(pxl)) for pxl in pxl_list]
return vals
def begin_interaction(type_, fnum):
print('[inter] starting %s interaction' % type_)
fig = df2.figure(fnum=fnum, docla=True, doclf=True)
ax = df2.gca()
df2.disconnect_callback(fig, 'button_press_event', axes=[ax])
return fig
def _sv_view(cx):
fnum = viz.FNUMS['special']
fig = df2.figure(fnum=fnum, docla=True, doclf=True)
df2.disconnect_callback(fig, 'button_press_event')
viz.viz_spatial_verification(hs, res.qcx, cx2=cx, fnum=fnum)
viz.draw()
def show_descriptors_match_distances(orgres2_distance,
fnum=1,
db_name='',
**kwargs):
disttype_list = orgres2_distance.itervalues().next().keys()
orgtype_list = orgres2_distance.keys()
(nRow, nCol) = len(orgtype_list), len(disttype_list)
nColors = nRow * nCol
color_list = df2.distinct_colors(nColors)
df2.figure(fnum=fnum, docla=True, doclf=True)
pnum_ = lambda px: (nRow, nCol, px + 1)
plot_type = helpers.get_arg('--plot-type', default='plot')
# Remember min and max val for each distance type (l1, emd...)
distkey2_min = {distkey: np.uint64(-1) for distkey in disttype_list}
distkey2_max = {distkey: 0 for distkey in disttype_list}
def _distplot(dists, color, label, distkey, plot_type=plot_type):
data = sorted(dists)
ax = df2.gca()
min_ = distkey2_min[distkey]
max_ = distkey2_max[distkey]
if plot_type == 'plot':
df2.plot(data, color=color, label=label)
#xticks = np.linspace(np.min(data), np.max(data), 3)
#yticks = np.linspace(0, len(data), 5)
#ax.set_xticks(xticks)
#ax.set_yticks(yticks)
ax.set_ylim(min_, max_)
ax.set_xlim(0, len(dists))
ax.set_ylabel('distance')
ax.set_xlabel('matches indexes (sorted by distance)')
df2.legend(loc='lower right')
if plot_type == 'pdf':
df2.plot_pdf(data, color=color, label=label)
ax.set_ylabel('pr')
ax.set_xlabel('distance')
ax.set_xlim(min_, max_)
df2.legend(loc='upper right')
df2.dark_background(ax)
df2.small_xticks(ax)
df2.small_yticks(ax)
px = 0
for orgkey in orgtype_list:
for distkey in disttype_list:
dists = orgres2_distance[orgkey][distkey]
if len(dists) == 0:
continue
min_ = dists.min()
max_ = dists.max()
distkey2_min[distkey] = min(distkey2_min[distkey], min_)
distkey2_max[distkey] = max(distkey2_max[distkey], max_)
for orgkey in orgtype_list:
for distkey in disttype_list:
print(((orgkey, distkey)))
dists = orgres2_distance[orgkey][distkey]
df2.figure(fnum=fnum, pnum=pnum_(px))
color = color_list[px]
title = distkey + ' ' + orgkey
label = 'P(%s | %s)' % (distkey, orgkey)
_distplot(dists, color, label, distkey, **kwargs)
#ax = df2.gca()
#ax.set_title(title)
px += 1
subtitle = 'the matching distances between sift descriptors'
title = '(sift) matching distances'
if db_name != '':
title = db_name + ' ' + title
df2.set_figtitle(title, subtitle)
df2.adjust_subplots_safe()
def _chip_view(pnum=(1, 1, 1), **kwargs):
df2.figure(fnum=fnum, pnum=pnum, docla=True, doclf=True)
# Toggle no keypoints view
viz.show_chip(hs, cx=cx, rchip=rchip, fnum=fnum, pnum=pnum, **kwargs)
df2.set_figtitle(figtitle)
def get_kp_border(rchip, kp):
np.set_printoptions(precision=8)
df2.reset()
df2.figure(9003, docla=True, doclf=True)
def _plotpts(data, px, color=df2.BLUE, label=''):
#df2.figure(9003, docla=True, pnum=(1, 1, px))
df2.plot2(data.T[0], data.T[1], '-', '', color=color, label=label)
df2.update()
def _plotarrow(x, y, dx, dy, color=df2.BLUE, label=''):
ax = df2.gca()
arrowargs = dict(head_width=.5, length_includes_head=True, label='')
arrow = df2.FancyArrow(x, y, dx, dy, **arrowargs)
arrow.set_edgecolor(color)
arrow.set_facecolor(color)
ax.add_patch(arrow)
df2.update()
def _2x2_eig(M2x2):
(evals, evecs) = np.linalg.eig(M2x2)
l1, l2 = evals
v1, v2 = evecs
return l1, l2, v1, v2
#-----------------------
# INPUT
#-----------------------
# We are given the keypoint in invA format
(x, y, ia11, ia21, ia22), ia12 = kp, 0
# invA2x2 is a transformation from points on a unit circle to the ellipse
invA2x2 = np.array([[ia11, ia12],
[ia21, ia22]])
#-----------------------
# DRAWING
#-----------------------
# Lets start off by drawing the ellipse that we are goign to work with
# Create unit circle sample
tau = 2 * np.pi
theta_list = np.linspace(0, tau, 1000)
cicrle_pts = np.array([(np.cos(t), np.sin(t)) for t in theta_list])
ellipse_pts = invA2x2.dot(cicrle_pts.T).T
_plotpts(ellipse_pts, 0, df2.BLACK, label='invA2x2.dot(unit_circle)')
l1, l2, v1, v2 = _2x2_eig(invA2x2)
dx1, dy1 = (v1 * l1)
dx2, dy2 = (v2 * l2)
_plotarrow(0, 0, dx1, dy1, color=df2.ORANGE, label='invA2x2 e1')
_plotarrow(0, 0, dx2, dy2, color=df2.RED, label='invA2x2 e2')
#-----------------------
# REPRESENTATION
#-----------------------
# A2x2 is a transformation from points on the ellipse to a unit circle
A2x2 = np.linalg.inv(invA2x2)
# Points on a matrix satisfy (x).T.dot(E2x2).dot(x) = 1
E2x2 = A2x2.T.dot(A2x2)
#Lets check our assertion: (x).T.dot(E2x2).dot(x) = 1
checks = [pt.T.dot(E2x2).dot(pt) for pt in ellipse_pts]
assert all([abs(1 - check) < 1E-11 for check in checks])
#-----------------------
# CONIC SECTIONS
#-----------------------
# All of this was from the Perdoch paper, now lets move into conic sections
# We will use the notation from wikipedia
# http://en.wikipedia.org/wiki/Conic_section
# http://en.wikipedia.org/wiki/Matrix_representation_of_conic_sections
# The matrix representation of a conic is:
((A, B, B_, C), (D, E, F)) = (E2x2.flatten(), (0, 0, 1))
assert B == B_, 'matrix should by symmetric'
# A_Q is our conic section (aka ellipse matrix)
A_Q = np.array((( A, B / 2, D / 2),
(B / 2, C, E / 2),
(D / 2, E / 2, F)))
assert np.linalg.det(A_Q) != 0, 'degenerate conic'
# As long as det(A_Q) is not 0 it is not degenerate and we can work with the
# minor 2x2 matrix
A_33 = np.array((( A, B / 2),
(B / 2, C)))
# (det == 0)->parabola, (det < 0)->hyperbola, (det > 0)->ellipse
assert np.linalg.det(A_33) > 0, 'conic is not an ellipse'
# Centers are obtained by solving for where the gradient of the quadratic
# becomes 0. Without going through the derivation the calculation is...
# These should be 0, 0 if we are at the origin, or our original x, y
# coordinate specified by the keypoints. I'm doing the calculation just for
# shits and giggles
x_center = (B * E - (2 * C * D)) / (4 * A * C - B ** 2)
y_center = (D * B - (2 * A * E)) / (4 * A * C - B ** 2)
#=================
# DRAWING
#=================
# Now we are going to determine the major and minor axis
# of this beast. It just the center augmented by the eigenvecs
l1, l2, v1, v2 = _2x2_eig(A_33)
dx1, dy1 = 0 - (v1 / np.sqrt(l1))
dx2, dy2 = 0 - (v2 / np.sqrt(l2))
_plotarrow(0, 0, dx1, dy1, color=df2.BLUE)
_plotarrow(0, 0, dx2, dy2, color=df2.BLUE)
# The angle between the major axis and our x axis is:
x_axis = np.array([1, 0])
theta = np.arccos(x_axis.dot(evec1))
# The eccentricity is determined by:
nu = 1
numer = 2 * np.sqrt((A - C) ** 2 + B ** 2)
denom = nu * (A + C) + np.sqrt((A - C) ** 2 + B ** 2)
eccentricity = np.sqrt(numer / denom)
from scipy.special import ellipeinc
# Algebraic form of connic
#assert (a * (x ** 2)) + (b * (x * y)) + (c * (y ** 2)) + (d * x) + (e * y) + (f) == 0
#---------------------
invA = np.array([[a, 0],
[c, d]])
Ashape = np.linalg.inv(np.array([[a, 0],
[c, d]]))
Ashape /= np.sqrt(np.linalg.det(Ashape))
tau = 2 * np.pi
nSamples = 100
theta_list = np.linspace(0, tau, nSamples)
# Create unit circle sample
cicrle_pts = np.array([(np.cos(t), np.sin(t)) for t in theta_list])
circle_hpts = np.hstack([cicrle_pts, np.ones((len(cicrle_pts), 1))])
# Transform as if the unit cirle was the warped patch
ashape_pts = Ashape.dot(cicrle_pts.T).T
inv = np.linalg.inv
svd = np.linalg.svd
U, S_, V = svd(Ashape)
S = np.diag(S_)
pxl_list3 = invA.dot(cicrle_pts[:, 0:2].T).T
pxl_list4 = invA.dot(ashape_pts[:, 0:2].T).T
pxl_list5 = invA.T.dot(cicrle_pts[:, 0:2].T).T
pxl_list6 = invA.T.dot(ashape_pts[:, 0:2].T).T
pxl_list7 = inv(V).dot(ashape_pts[:, 0:2].T).T
pxl_list8 = inv(U).dot(ashape_pts[:, 0:2].T).T
df2.draw()
def _plot(data, px, title=''):
df2.figure(9003, docla=True, pnum=(2, 4, px))
df2.plot2(data.T[0], data.T[1], '.', title)
df2.figure(9003, doclf=True)
_plot(cicrle_pts, 1, 'unit circle')
_plot(ashape_pts, 2, 'A => circle shape')
_plot(pxl_list3, 3)
_plot(pxl_list4, 4)
_plot(pxl_list5, 5)
_plot(pxl_list6, 6)
_plot(pxl_list7, 7)
_plot(pxl_list8, 8)
df2.draw()
invA = np.array([[a, 0, x],
[c, d, y],
[0, 0, 1]])
pxl_list = invA.dot(circle_hpts.T).T[:, 0:2]
df2.figure(9002, doclf=True)
df2.imshow(rchip)
df2.plot2(pxl_list.T[0], pxl_list.T[1], '.')
df2.draw()
vals = [cv2.getRectSubPix(rchip, (1, 1), tuple(pxl)) for pxl in pxl_list]
return vals
def show_descriptors_match_distances(orgres2_distance, fnum=1, db_name='', **kwargs):
disttype_list = orgres2_distance.itervalues().next().keys()
orgtype_list = orgres2_distance.keys()
(nRow, nCol) = len(orgtype_list), len(disttype_list)
nColors = nRow * nCol
color_list = df2.distinct_colors(nColors)
df2.figure(fnum=fnum, docla=True, doclf=True)
pnum_ = lambda px: (nRow, nCol, px + 1)
plot_type = helpers.get_arg('--plot-type', default='plot')
# Remember min and max val for each distance type (l1, emd...)
distkey2_min = {distkey: np.uint64(-1) for distkey in disttype_list}
distkey2_max = {distkey: 0 for distkey in disttype_list}
def _distplot(dists, color, label, distkey, plot_type=plot_type):
data = sorted(dists)
ax = df2.gca()
min_ = distkey2_min[distkey]
max_ = distkey2_max[distkey]
if plot_type == 'plot':
df2.plot(data, color=color, label=label)
#xticks = np.linspace(np.min(data), np.max(data), 3)
#yticks = np.linspace(0, len(data), 5)
#ax.set_xticks(xticks)
#ax.set_yticks(yticks)
ax.set_ylim(min_, max_)
ax.set_xlim(0, len(dists))
ax.set_ylabel('distance')
ax.set_xlabel('matches indexes (sorted by distance)')
df2.legend(loc='lower right')
if plot_type == 'pdf':
df2.plot_pdf(data, color=color, label=label)
ax.set_ylabel('pr')
ax.set_xlabel('distance')
ax.set_xlim(min_, max_)
df2.legend(loc='upper right')
df2.dark_background(ax)
df2.small_xticks(ax)
df2.small_yticks(ax)
px = 0
for orgkey in orgtype_list:
for distkey in disttype_list:
dists = orgres2_distance[orgkey][distkey]
if len(dists) == 0:
continue
min_ = dists.min()
max_ = dists.max()
distkey2_min[distkey] = min(distkey2_min[distkey], min_)
distkey2_max[distkey] = max(distkey2_max[distkey], max_)
for orgkey in orgtype_list:
for distkey in disttype_list:
print(((orgkey, distkey)))
dists = orgres2_distance[orgkey][distkey]
df2.figure(fnum=fnum, pnum=pnum_(px))
color = color_list[px]
title = distkey + ' ' + orgkey
label = 'P(%s | %s)' % (distkey, orgkey)
_distplot(dists, color, label, distkey, **kwargs)
#ax = df2.gca()
#ax.set_title(title)
px += 1
subtitle = 'the matching distances between sift descriptors'
title = '(sift) matching distances'
if db_name != '':
title = db_name + ' ' + title
df2.set_figtitle(title, subtitle)
df2.adjust_subplots_safe()