def compute_target_size_from_aspect(chip_list, target_height=256):
r"""
Args:
chip_list (list):
target_height (int):
Returns:
?: data
CommandLine:
python -m ibeis_cnn.ibsplugin --test-compute_target_size_from_aspect
Example:
>>> # DISABLE_DOCTEST
>>> from ibeis_cnn.ibsplugin import * # NOQA
>>> import ibeis
>>> # build test data
>>> ibs = ibeis.opendb('testdb1')
>>> (aid1_list, aid2_list) = get_identify_training_aid_pairs(ibs)
>>> # execute function
>>> chip_list = ibs.get_annot_chips(ut.unique_ordered2(aid1_list + aid2_list))
>>> target_height = 256
>>> target_size = compute_target_size_from_aspect(chip_list, target_height)
>>> # verify results
>>> result = str(target_size)
>>> print(result)
(381, 256)
"""
import vtool as vt
sizes1 = np.array([vt.get_size(chip1) for chip1 in chip_list])
ar1_list = sizes1.T[0] / sizes1.T[1]
ave_ar = ar1_list.mean()
target_size = (int(np.round(ave_ar * target_height)), target_height)
return target_size
def dummy_preproc_chip(depc, annot_rowid_list, config=None):
"""
TODO: Infer properties from docstr?
Args:
depc (dtool.DependencyCache):
annot_rowid_list (list): list of annot rowids
config (dict): config dictionary
Returns:
tuple : ((int, int), ('extern', vt.imread))
"""
if config is None:
config = {}
# Demonstates using asobject to get input to function as a dictionary
# of properties
#for annot in annot_list:
#print('[preproc] Computing chips of aid=%r' % (aid,))
print('[preproc] Computing chips')
for aid in annot_rowid_list:
#aid = annot['aid']
#chip_fpath = annot['gpath']
chip_fpath = gpath_list[aid]
#w, h = vt.image.open_image_size(chip_fpath)
chip = vt.imread(chip_fpath)
size = vt.get_size(chip)
#size = (w, h)
#print('* chip_fpath = %r' % (chip_fpath,))
#print('* size = %r' % (size,))
#yield size, chip_fpath
yield size, chip
def dummy_preproc_chip(depc, annot_rowid_list, config=None):
"""
TODO: Infer properties from docstr?
Args:
depc (dtool.DependencyCache):
annot_rowid_list (list): list of annot rowids
config (dict): config dictionary
Returns:
tuple : ((int, int), ('extern', vt.imread))
"""
if config is None:
config = {}
# Demonstates using asobject to get input to function as a dictionary
# of properties
#for annot in annot_list:
#print('[preproc] Computing chips of aid=%r' % (aid,))
print('[preproc] Computing chips')
for aid in annot_rowid_list:
#aid = annot['aid']
#chip_fpath = annot['gpath']
chip_fpath = gpath_list[aid]
#w, h = vt.image.open_image_size(chip_fpath)
chip = vt.imread(chip_fpath)
size = vt.get_size(chip)
#size = (w, h)
print('Dummpy preproc chip yeilds')
print('* chip_fpath = %r' % (chip_fpath, ))
print('* size = %r' % (size, ))
#yield size, chip_fpath
yield size, chip
def get_blended_chip(chip1, chip2, M):
"""
warps chip1 into chip2 space
"""
wh2 = vt.get_size(chip2)
chip1_Mt = vt.warpHomog(chip1, M, wh2)
chip2_blendM = vt.blend_images(chip1_Mt, chip2)
return chip2_blendM
def get_blended_chip(chip1, chip2, M):
"""
warps chip1 into chip2 space
"""
wh2 = vt.get_size(chip2)
chip1_Mt = vt.warpHomog(chip1, M, wh2)
chip2_blendM = vt.blend_images(chip1_Mt, chip2)
return chip2_blendM
def dummy_manual_chipmask(depc, parent_rowids, config=None):
import vtool as vt
from plottool import interact_impaint
mask_dpath = join(depc.cache_dpath, 'ManualChipMask')
ut.ensuredir(mask_dpath)
if config is None:
config = {}
print('Requesting user defined chip mask')
for rowid in parent_rowids:
img = vt.imread(gpath_list[rowid])
mask = interact_impaint.impaint_mask2(img)
mask_fpath = join(mask_dpath, 'mask%d.png' % (rowid, ))
vt.imwrite(mask_fpath, mask)
w, h = vt.get_size(mask)
yield (w, h), mask_fpath
def dummy_manual_chipmask(depc, parent_rowids, config=None):
import vtool as vt
from plottool import interact_impaint
mask_dpath = ut.unixjoin(depc.cache_dpath, 'ManualChipMask')
ut.ensuredir(mask_dpath)
if config is None:
config = {}
print('Requesting user defined chip mask')
for rowid in parent_rowids:
img = vt.imread(gpath_list[rowid])
mask = interact_impaint.impaint_mask2(img)
mask_fpath = ut.unixjoin(mask_dpath, 'mask%d.png' % (rowid,))
vt.imwrite(mask_fpath, mask)
w, h = vt.get_size(mask)
yield (w, h), mask_fpath
def makewarp(imgBGR):
# hack a projection matrix using dummy homogrpahy
imgBGRA = cv2.cvtColor(imgBGR, cv2.COLOR_BGR2BGRA)
imgBGRA[:, :, 3] = .87 * 255 # hack alpha
imgBGRA = vt.pad_image(imgBGRA, 2, value=[0, 0, 255, 255])
size = np.array(vt.get_size(imgBGRA))
pts1 = np.array([(0, 0), (0, 1), (1, 1), (1, 0)]) * size
x_adjust = .15
y_adjust = .5
pts2 = np.array([(x_adjust, 0), (0, 1 - y_adjust), (1, 1 - y_adjust), (1 - x_adjust, 0)]) * size
H = cv2.findHomography(pts1, pts2)[0]
dsize = np.array(vt.bbox_from_verts(pts2)[2:4]).astype(np.int)
warpkw = dict(flags=cv2.INTER_LANCZOS4, borderMode=cv2.BORDER_CONSTANT)
imgBGRA_warped = cv2.warpPerspective(imgBGRA, H, tuple(dsize), **warpkw)
return imgBGRA_warped
def nosql_draw2(check_func, match):
from matplotlib.backends.backend_agg import FigureCanvas
try:
from matplotlib.backends.backend_agg import Figure
except ImportError:
from matplotlib.figure import Figure
was_interactive = mpl.is_interactive()
if was_interactive:
mpl.interactive(False)
# fnum = 32
fig = Figure()
canvas = FigureCanvas(fig) # NOQA
# fig.clf()
ax = fig.add_subplot(1, 1, 1)
if check_func is not None and check_func():
return
ax, xywh1, xywh2 = match.show(ax=ax)
if check_func is not None and check_func():
return
savekw = {
# 'dpi' : 60,
'dpi': 80,
}
axes_extents = pt.extract_axes_extents(fig)
# assert len(axes_extents) == 1, 'more than one axes'
extent = axes_extents[0]
with io.BytesIO() as stream:
# This call takes 23% - 15% of the time depending on settings
fig.savefig(stream, bbox_inches=extent, **savekw)
stream.seek(0)
data = np.fromstring(stream.getvalue(), dtype=np.uint8)
if check_func is not None and check_func():
return
pt.plt.close(fig)
image = cv2.imdecode(data, 1)
thumbsize = 221
max_dsize = (thumbsize, thumbsize)
dsize, sx, sy = vt.resized_clamped_thumb_dims(
vt.get_size(image), max_dsize)
if check_func is not None and check_func():
return
image = vt.resize(image, dsize)
vt.imwrite(fpath, image)
if check_func is not None and check_func():
return
def makewarp(imgBGR):
# hack a projection matrix using dummy homogrpahy
imgBGRA = cv2.cvtColor(imgBGR, cv2.COLOR_BGR2BGRA)
imgBGRA[:, :, 3] = .87 * 255 # hack alpha
imgBGRA = vt.pad_image(imgBGRA, 2, value=[0, 0, 255, 255])
size = np.array(vt.get_size(imgBGRA))
pts1 = np.array([(0, 0), (0, 1), (1, 1), (1, 0)]) * size
x_adjust = .15
y_adjust = .5
pts2 = np.array([(x_adjust, 0), (0, 1 - y_adjust), (1, 1 - y_adjust),
(1 - x_adjust, 0)]) * size
H = cv2.findHomography(pts1, pts2)[0]
dsize = np.array(vt.bbox_from_verts(pts2)[2:4]).astype(np.int)
warpkw = dict(flags=cv2.INTER_LANCZOS4, borderMode=cv2.BORDER_CONSTANT)
imgBGRA_warped = cv2.warpPerspective(imgBGRA, H, tuple(dsize),
**warpkw)
return imgBGRA_warped
def compute_tips(depc, chip_rowids, config=None):
manual_extract = config['manual_extract']
chips = depc.get_native('chip', chip_rowids, 'img')
for chip in chips:
seed = (chip).sum()
perb = ((seed % 1000) / 1000) * .25
w, h = vt.get_size(chip)
if manual_extract:
# Make noticable difference between config outputs
lpb = np.ceil(w * perb)
npb = np.ceil(h * perb)
rpb = -np.ceil(w * perb)
else:
lpb = np.ceil(w * perb / 2)
npb = -np.ceil(h * perb)
rpb = -np.ceil(w * perb)
wh = np.array([w, h], dtype=np.int32)[None, :]
rel_base = np.array([[.0, .5], [.5, .5], [1., .5]])
offset = np.array([[lpb, 0], [0, npb], [rpb, 0]])
tip = np.round((wh * rel_base)) + offset
left, notch, right = tip
yield left, notch, right
def compute_tips(depc, chip_rowids, config=None):
manual_extract = config['manual_extract']
chips = depc.get_native('chip', chip_rowids, 'img')
for chip in chips:
seed = (chip).sum()
perb = ((seed % 1000) / 1000) * .25
w, h = vt.get_size(chip)
if manual_extract:
# Make noticable difference between config outputs
lpb = np.ceil(w * perb)
npb = np.ceil(h * perb)
rpb = -np.ceil(w * perb)
else:
lpb = np.ceil(w * perb / 2)
npb = -np.ceil(h * perb)
rpb = -np.ceil(w * perb)
wh = np.array([w, h], dtype=np.int32)[None, :]
rel_base = np.array([[.0, .5], [.5, .5], [1., .5]])
offset = np.array([[lpb, 0], [0, npb], [rpb, 0]])
tip = np.round((wh * rel_base)) + offset
left, notch, right = tip
yield left, notch, right
def show_sv(chip1,
chip2,
kpts1,
kpts2,
fm,
homog_tup=None,
aff_tup=None,
mx=None,
show_assign=True,
show_lines=True,
show_kpts=True,
show_aff=None,
fnum=1,
refine_method=None,
**kwargs):
""" Visualizes spatial verification
CommandLine:
python -m vtool.spatial_verification --test-spatially_verify_kpts --show
"""
import vtool as vt
# import wbia.plottool as pt
# GEt Matching chips
kpts1_m = kpts1[fm.T[0]]
kpts2_m = kpts2[fm.T[1]]
wh2 = vt.get_size(chip2)
#
# Get Affine Chips, Keypoints, Inliers
if show_aff is None:
show_aff_ = aff_tup is not None
else:
show_aff_ = show_aff
if show_aff_:
(aff_inliers, Aff) = aff_tup
chip1_At = vt.warpAffine(chip1, Aff, wh2)
# kpts1_mAt = ktool.transform_kpts(kpts1_m, Aff)
chip2_blendA = vt.blend_images(chip1_At, chip2)
#
# Get Homog Chips, Keypoints, Inliers
show_homog = homog_tup is not None
if show_homog:
(hom_inliers, H) = homog_tup
# kpts1_mHt = ktool.transform_kpts(kpts1_m, H)
chip1_Ht = vt.warpHomog(chip1, H, wh2)
chip2_blendH = vt.blend_images(chip1_Ht, chip2)
#
# Drawing settings
nRows = (show_assign) + (show_aff_) + (show_homog)
nCols1 = (show_assign) + (show_aff_) + (show_homog)
nCols2 = 2
pnum1_ = df2.get_pnum_func(nRows, nCols1)
pnum2_ = df2.get_pnum_func(nRows, nCols2)
# in_kwargs = dict(rect=True, ell_alpha=.7, eig=False, ori=True, pts=True)
# out_kwargs = dict(rect=False, ell_alpha=.3, eig=False)
in_kwargs = dict(rect=False,
ell_alpha=0.7,
eig=False,
ori=False,
pts=False)
out_kwargs = dict(rect=False, ell_alpha=0.7, eig=False, ori=False)
def _draw_kpts(*args, **kwargs):
if not show_kpts:
return
df2.draw_kpts2(*args, **kwargs)
def draw_inlier_kpts(kpts_m, inliers, color, H=None):
_draw_kpts(kpts_m[inliers], color=color, H=H, **in_kwargs)
if mx is not None:
_draw_kpts(kpts_m[mx:(mx + 1)],
color=color,
ell_linewidth=3,
H=H,
**in_kwargs)
def _draw_matches(px, title, inliers):
dmkwargs = dict(
fs=None,
title=title,
all_kpts=False,
draw_lines=True,
docla=True,
draw_border=True,
fnum=fnum,
pnum=pnum1_(px),
colors=df2.ORANGE,
)
__fm = np.vstack((inliers, inliers)).T
df2.show_chipmatch2(chip1, chip2, kpts1_m, kpts2_m, __fm, **dmkwargs)
return px + 1
from wbia.plottool import color_funcs
colors = df2.distinct_colors(2, brightness=0.95)
color1, color2 = colors[0], colors[1]
color1_dark = color_funcs.darken_rgb(color1, 0.2)
color2_dark = color_funcs.darken_rgb(color2, 0.2)
def _draw_chip(px, title, chip, inliers, kpts1_m, kpts2_m, H1=None):
if isinstance(px, tuple):
pnum = px
df2.imshow(chip, title=title, fnum=fnum, pnum=pnum)
px = pnum[2]
else:
df2.imshow(chip, title=title, fnum=fnum, pnum=pnum2_(px))
if kpts1_m is not None:
_draw_kpts(kpts1_m, color=color1_dark, H=H1, **out_kwargs)
draw_inlier_kpts(kpts1_m, inliers, color1, H=H1)
if kpts2_m is not None:
_draw_kpts(kpts2_m, color=color2_dark, **out_kwargs)
draw_inlier_kpts(kpts2_m, inliers, color2)
if kpts2_m is not None and kpts1_m is not None and show_lines:
__fm = np.vstack((inliers, inliers)).T
df2.draw_lines2(
kpts1_m,
kpts2_m,
__fm,
color_list=[custom_constants.ORANGE],
lw=2,
line_alpha=1,
H1=H1,
)
return px + 1
#
# Begin the drawing
df2.figure(fnum=fnum, pnum=(nRows, nCols1, 1), docla=True, doclf=True)
px = 0
if show_assign:
# Draw the Assigned -> Affine -> Homography matches
px = _draw_matches(px, '%d Assigned matches ' % len(fm),
np.arange(len(fm)))
if show_aff_:
px = _draw_matches(px, '%d Initial inliers ' % len(aff_inliers),
aff_inliers)
if show_homog:
if refine_method is None:
refine_method = ''
if len(refine_method) > 0:
refine_method_ = '(%s) ' % (refine_method, )
else:
refine_method_ = ''
px = _draw_matches(
px,
'%d Refined %sinliers' % (len(hom_inliers), refine_method_),
hom_inliers,
)
#
# Draw the Affine Transformations
px = nCols2 * show_assign
# if show_aff_ or show_homog:
if show_aff_:
# px = _draw_chip(px, 'Source', chip1, aff_inliers, kpts1_m, None)
# px = _draw_chip(px, 'Dest', chip2, aff_inliers, None, kpts2_m)
px = _draw_chip(px,
'Initial Warped',
chip1_At,
aff_inliers,
kpts1_m,
None,
H1=Aff)
px = _draw_chip(px,
'Initial Blend',
chip2_blendA,
aff_inliers,
kpts1_m,
kpts2_m,
H1=Aff)
# px = _draw_chip(px, 'Affine', chip1_At, aff_inliers, kpts1_mAt, None)
# px = _draw_chip(px, 'Aff Blend', chip2_blendA, aff_inliers, kpts1_mAt, kpts2_m)
pass
#
# Draw the Homography Transformation
if show_homog:
# px = _draw_chip(px, 'Source', chip1, hom_inliers, kpts1_m, None)
# px = _draw_chip(px, 'Dest', chip2, hom_inliers, None, kpts2_m)
# px = _draw_chip(px, 'Homog', chip1_Ht, hom_inliers, kpts1_mHt, None)
# px = _draw_chip(px, 'Homog Blend', chip2_blendH, hom_inliers, kpts1_mHt, kpts2_m)
px = _draw_chip(px,
'Refined Warped',
chip1_Ht,
hom_inliers,
kpts1_m,
None,
H1=H)
px = _draw_chip(px,
'Refined Blend',
chip2_blendH,
hom_inliers,
kpts1_m,
kpts2_m,
H1=H)
def test_mser():
import cv2
import vtool as vt
import plottool as pt
import numpy as np
pt.qt4ensure()
class Keypoints(ut.NiceRepr):
"""
Convinence class for dealing with keypoints
"""
def __init__(self, kparr, info=None):
self.kparr = kparr
if info is None:
info = {}
self.info = info
def add_info(self, key, val):
self.info[key] = val
def __nice__(self):
return ' ' + str(len(self.kparr))
@property
def scale(self):
return vt.get_scales(self.kparr)
@property
def eccentricity(self):
return vt.get_kpts_eccentricity(self.kparr)
def compress(self, flags, inplace=False):
subarr = self.kparr.compress(flags, axis=0)
info = {
key: ut.compress(val, flags)
for key, val in self.info.items()
}
return Keypoints(subarr, info)
img_fpath = ut.grab_test_imgpath(
ut.get_argval('--fname', default='zebra.png'))
imgBGR = vt.imread(img_fpath)
imgGray = cv2.cvtColor(imgBGR, cv2.COLOR_BGR2GRAY)
# http://docs.opencv.org/master/d3/d28/classcv_1_1MSER.html#gsc.tab=0
# http://stackoverflow.com/questions/17647500/exact-meaning-of-the-parameters-given-to-initialize-mser-in-opencv-2-4-x
factory = cv2.MSER_create
img_area = np.product(np.array(vt.get_size(imgGray)))
_max_area = (img_area // 10)
_delta = 8
_min_diversity = .5
extractor = factory(_delta=_delta,
_max_area=_max_area,
_min_diversity=_min_diversity)
# bboxes are x,y,w,h
regions, bboxes = extractor.detectRegions(imgGray)
# ellipse definition from [Fitzgibbon95]
# http://www.bmva.org/bmvc/1995/bmvc-95-050.pdf p518
# ell = [c_x, c_y, R_x, R_y, theta]
# (cx, cy) = conic center
# Rx and Ry = conic radii
# theta is the counterclockwise angle
fitz_ellipses = [cv2.fitEllipse(mser) for mser in regions]
# http://answers.opencv.org/question/19015/how-to-use-mser-in-python/
#hulls = [cv2.convexHull(p.reshape(-1, 1, 2)) for p in regions]
#hull_ells = [cv2.fitEllipse(hull[:, 0]) for hull in hulls]
invVR_mats = []
for ell in fitz_ellipses:
((cx, cy), (dx, dy), degrees) = ell
theta = np.radians(degrees) # opencv lives in radians
# Convert diameter to radians
rx = dx / 2
ry = dy / 2
S = vt.scale_mat3x3(rx, ry)
T = vt.translation_mat3x3(cx, cy)
R = vt.rotation_mat3x3(theta)
invVR = T.dot(R.dot(S))
invVR_mats.append(invVR)
invVR_mats = np.array(invVR_mats)
#_oris = vt.get_invVR_mats_oris(invVR_mats)
kpts2_ = vt.flatten_invV_mats_to_kpts(invVR_mats)
self = Keypoints(kpts2_)
self.add_info('regions', regions)
flags = (self.eccentricity < .9)
#flags = self.scale < np.mean(self.scale)
#flags = self.scale < np.median(self.scale)
self = self.compress(flags)
import plottool as pt
#pt.interact_keypoints.ishow_keypoints(imgBGR, self.kparr, None, ell_alpha=.4, color='distinct', fnum=2)
#import plottool as pt
vis = imgBGR.copy()
for region in self.info['regions']:
vis[region.T[1], region.T[0], :] = 0
#regions, bbox = mser.detectRegions(gray)
#hulls = [cv2.convexHull(p.reshape(-1, 1, 2)) for p in self.info['regions']]
#cv2.polylines(vis, hulls, 1, (0, 255, 0))
#for region in self.info['regions']:
# ell = cv2.fitEllipse(region)
# cv2.ellipse(vis, ell, (255))
pt.interact_keypoints.ishow_keypoints(vis,
self.kparr,
None,
ell_alpha=.4,
color='distinct',
fnum=2)
#pt.imshow(vis, fnum=2)
pt.update()
#extractor = extract_factory['DAISY']()
#desc_type_to_dtype = {
# cv2.CV_8U: np.uint8,
# cv2.CV_8s: np.uint,
#}
#def alloc_desc(extractor):
# desc_type = extractor.descriptorType()
# desc_size = extractor.descriptorSize()
# dtype = desc_type_to_dtype[desc_type]
# shape = (len(cv2_kpts), desc_size)
# desc = np.empty(shape, dtype=dtype)
# return desc
#ut.search_module(cv2, 'array', recursive=True)
#ut.search_module(cv2, 'freak', recursive=True)
#ut.search_module(cv2, 'out', recursive=True)
#cv2_kpts = cv2_kpts[0:2]
#for key, factory in just_desc_factory_.items():
# extractor = factory()
# desc = alloc_desc(extractor)
# desc = extractor.compute(imgGray, cv2_kpts)
# feats[key] = (desc,)
# #extractor.compute(imgGray, cv2_kpts, desc)
# pass
#kpts = np.array(list(map(from_cv2_kpts, cv2_kpts)))
#orb = cv2.ORB()
#kp1, des1 = orb.detectAndCompute(imgGray, None)
#blober = cv2.SimpleBlobDetector_create()
#haris_kpts = cv2.cornerHarris(imgGray, 2, 3, 0.04)
#[name for name in dir(cv2) if 'mat' in name.lower()]
#[name for name in dir(cv2.xfeatures2d) if 'desc' in name.lower()]
#[name for name in dir(cv2) if 'detect' in name.lower()]
#[name for name in dir(cv2) if 'extract' in name.lower()]
#[name for name in dir(cv2) if 'ellip' in name.lower()]
#sift = cv2.xfeatures2d.SIFT_create()
#cv2_kpts = sift.detect(imgGray)
#desc = sift.compute(imgGray, cv2_kpts)[1]
#freak = cv2.xfeatures2d.FREAK_create()
#cv2_kpts = freak.detect(imgGray)
#desc = freak.compute(imgGray, cv2_kpts)[1]
pass
def compute_chip(depc, aids, config=None):
for aid in aids:
chip = fontdemo.get_text_test_img(str(aid))
size = vt.get_size(chip)
yield size, chip
def test_mser():
import cv2
import vtool as vt
import plottool as pt
import numpy as np
pt.qt4ensure()
class Keypoints(ut.NiceRepr):
"""
Convinence class for dealing with keypoints
"""
def __init__(self, kparr, info=None):
self.kparr = kparr
if info is None:
info = {}
self.info = info
def add_info(self, key, val):
self.info[key] = val
def __nice__(self):
return ' ' + str(len(self.kparr))
@property
def scale(self):
return vt.get_scales(self.kparr)
@property
def eccentricity(self):
return vt.get_kpts_eccentricity(self.kparr)
def compress(self, flags, inplace=False):
subarr = self.kparr.compress(flags, axis=0)
info = {key: ut.compress(val, flags) for key, val in self.info.items()}
return Keypoints(subarr, info)
img_fpath = ut.grab_test_imgpath(ut.get_argval('--fname', default='zebra.png'))
imgBGR = vt.imread(img_fpath)
imgGray = cv2.cvtColor(imgBGR, cv2.COLOR_BGR2GRAY)
# http://docs.opencv.org/master/d3/d28/classcv_1_1MSER.html#gsc.tab=0
# http://stackoverflow.com/questions/17647500/exact-meaning-of-the-parameters-given-to-initialize-mser-in-opencv-2-4-x
factory = cv2.MSER_create
img_area = np.product(np.array(vt.get_size(imgGray)))
_max_area = (img_area // 10)
_delta = 8
_min_diversity = .5
extractor = factory(_delta=_delta, _max_area=_max_area, _min_diversity=_min_diversity)
# bboxes are x,y,w,h
regions, bboxes = extractor.detectRegions(imgGray)
# ellipse definition from [Fitzgibbon95]
# http://www.bmva.org/bmvc/1995/bmvc-95-050.pdf p518
# ell = [c_x, c_y, R_x, R_y, theta]
# (cx, cy) = conic center
# Rx and Ry = conic radii
# theta is the counterclockwise angle
fitz_ellipses = [cv2.fitEllipse(mser) for mser in regions]
# http://answers.opencv.org/question/19015/how-to-use-mser-in-python/
#hulls = [cv2.convexHull(p.reshape(-1, 1, 2)) for p in regions]
#hull_ells = [cv2.fitEllipse(hull[:, 0]) for hull in hulls]
invVR_mats = []
for ell in fitz_ellipses:
((cx, cy), (dx, dy), degrees) = ell
theta = np.radians(degrees) # opencv lives in radians
# Convert diameter to radians
rx = dx / 2
ry = dy / 2
S = vt.scale_mat3x3(rx, ry)
T = vt.translation_mat3x3(cx, cy)
R = vt.rotation_mat3x3(theta)
invVR = T.dot(R.dot(S))
invVR_mats.append(invVR)
invVR_mats = np.array(invVR_mats)
#_oris = vt.get_invVR_mats_oris(invVR_mats)
kpts2_ = vt.flatten_invV_mats_to_kpts(invVR_mats)
self = Keypoints(kpts2_)
self.add_info('regions', regions)
flags = (self.eccentricity < .9)
#flags = self.scale < np.mean(self.scale)
#flags = self.scale < np.median(self.scale)
self = self.compress(flags)
import plottool as pt
#pt.interact_keypoints.ishow_keypoints(imgBGR, self.kparr, None, ell_alpha=.4, color='distinct', fnum=2)
#import plottool as pt
vis = imgBGR.copy()
for region in self.info['regions']:
vis[region.T[1], region.T[0], :] = 0
#regions, bbox = mser.detectRegions(gray)
#hulls = [cv2.convexHull(p.reshape(-1, 1, 2)) for p in self.info['regions']]
#cv2.polylines(vis, hulls, 1, (0, 255, 0))
#for region in self.info['regions']:
# ell = cv2.fitEllipse(region)
# cv2.ellipse(vis, ell, (255))
pt.interact_keypoints.ishow_keypoints(vis, self.kparr, None, ell_alpha=.4, color='distinct', fnum=2)
#pt.imshow(vis, fnum=2)
pt.update()
#extractor = extract_factory['DAISY']()
#desc_type_to_dtype = {
# cv2.CV_8U: np.uint8,
# cv2.CV_8s: np.uint,
#}
#def alloc_desc(extractor):
# desc_type = extractor.descriptorType()
# desc_size = extractor.descriptorSize()
# dtype = desc_type_to_dtype[desc_type]
# shape = (len(cv2_kpts), desc_size)
# desc = np.empty(shape, dtype=dtype)
# return desc
#ut.search_module(cv2, 'array', recursive=True)
#ut.search_module(cv2, 'freak', recursive=True)
#ut.search_module(cv2, 'out', recursive=True)
#cv2_kpts = cv2_kpts[0:2]
#for key, factory in just_desc_factory_.items():
# extractor = factory()
# desc = alloc_desc(extractor)
# desc = extractor.compute(imgGray, cv2_kpts)
# feats[key] = (desc,)
# #extractor.compute(imgGray, cv2_kpts, desc)
# pass
#kpts = np.array(list(map(from_cv2_kpts, cv2_kpts)))
#orb = cv2.ORB()
#kp1, des1 = orb.detectAndCompute(imgGray, None)
#blober = cv2.SimpleBlobDetector_create()
#haris_kpts = cv2.cornerHarris(imgGray, 2, 3, 0.04)
#[name for name in dir(cv2) if 'mat' in name.lower()]
#[name for name in dir(cv2.xfeatures2d) if 'desc' in name.lower()]
#[name for name in dir(cv2) if 'detect' in name.lower()]
#[name for name in dir(cv2) if 'extract' in name.lower()]
#[name for name in dir(cv2) if 'ellip' in name.lower()]
#sift = cv2.xfeatures2d.SIFT_create()
#cv2_kpts = sift.detect(imgGray)
#desc = sift.compute(imgGray, cv2_kpts)[1]
#freak = cv2.xfeatures2d.FREAK_create()
#cv2_kpts = freak.detect(imgGray)
#desc = freak.compute(imgGray, cv2_kpts)[1]
pass
def nosql_draw(check_func, rchip1_fpath, rchip2_fpath, kpts1, kpts2):
# This gets executed in the child thread and does drawing async style
#from matplotlib.backends.backend_pdf import FigureCanvasPdf as FigureCanvas
#from matplotlib.backends.backend_pdf import Figure
#from matplotlib.backends.backend_svg import FigureCanvas
#from matplotlib.backends.backend_svg import Figure
from matplotlib.backends.backend_agg import FigureCanvas
try:
from matplotlib.backends.backend_agg import Figure
except ImportError:
from matplotlib.figure import Figure
kpts1_ = vt.offset_kpts(kpts1, (0, 0), (resize_factor, resize_factor))
kpts2_ = vt.offset_kpts(kpts2, (0, 0), (resize_factor, resize_factor))
#from matplotlib.figure import Figure
if check_func is not None and check_func():
return
rchip1 = vt.imread(rchip1_fpath)
rchip1 = vt.resize_image_by_scale(rchip1, resize_factor)
if check_func is not None and check_func():
return
rchip2 = vt.imread(rchip2_fpath)
rchip2 = vt.resize_image_by_scale(rchip2, resize_factor)
if check_func is not None and check_func():
return
try:
idx = cm.daid2_idx[daid]
fm = cm.fm_list[idx]
fsv = None if cm.fsv_list is None else cm.fsv_list[idx]
fs = None if fsv is None else fsv.prod(axis=1)
except KeyError:
fm = []
fs = None
fsv = None
maxnum = 200
if fs is not None and len(fs) > maxnum:
# HACK TO ONLY SHOW TOP MATCHES
sortx = fs.argsort()[::-1]
fm = fm.take(sortx[:maxnum], axis=0)
fs = fs.take(sortx[:maxnum], axis=0)
was_interactive = mpl.is_interactive()
if was_interactive:
mpl.interactive(False)
#fnum = 32
fig = Figure()
canvas = FigureCanvas(fig) # NOQA
#fig.clf()
ax = fig.add_subplot(1, 1, 1)
if check_func is not None and check_func():
return
#fig = pt.plt.figure(fnum)
#H1 = np.eye(3)
#H2 = np.eye(3)
#H1[0, 0] = .5
#H1[1, 1] = .5
#H2[0, 0] = .5
#H2[1, 1] = .5
ax, xywh1, xywh2 = pt.show_chipmatch2(rchip1, rchip2, kpts1_, kpts2_, fm,
fs=fs, colorbar_=False, ax=ax)
if check_func is not None and check_func():
return
savekw = {
# 'dpi' : 60,
'dpi' : 80,
}
axes_extents = pt.extract_axes_extents(fig)
#assert len(axes_extents) == 1, 'more than one axes'
extent = axes_extents[0]
with io.BytesIO() as stream:
# This call takes 23% - 15% of the time depending on settings
fig.savefig(stream, bbox_inches=extent, **savekw)
stream.seek(0)
data = np.fromstring(stream.getvalue(), dtype=np.uint8)
if check_func is not None and check_func():
return
pt.plt.close(fig)
image = cv2.imdecode(data, 1)
thumbsize = 221
max_dsize = (thumbsize, thumbsize)
dsize, sx, sy = vt.resized_clamped_thumb_dims(vt.get_size(image), max_dsize)
if check_func is not None and check_func():
return
image = vt.resize(image, dsize)
vt.imwrite(fpath, image)
if check_func is not None and check_func():
return
def compute_chip(depc, aids, config=None):
for aid in aids:
chip = fontdemo.get_text_test_img(str(aid))
size = vt.get_size(chip)
yield size, chip
def netx_draw_images_at_positions(img_list, pos_list, size_list, color_list,
framewidth_list):
"""
Overlays images on a networkx graph
References:
https://gist.github.com/shobhit/3236373
http://matplotlib.org/examples/pylab_examples/demo_annotation_box.html
http://stackoverflow.com/questions/11487797/mpl-overlay-small-image
http://matplotlib.org/api/text_api.html
http://matplotlib.org/api/offsetbox_api.html
TODO: look into DraggableAnnotation
"""
#print('[viz_graph] drawing %d images' % len(img_list))
# Thumb stackartist
import plottool as pt
#ax = pt.gca()
# Ensure all images have been read
img_list_ = [vt.convert_colorspace(vt.imread(img), 'RGB')
if isinstance(img, six.string_types) else img
for img in img_list]
size_list_ = [vt.get_size(img) if size is None else size
for size, img in zip(size_list, img_list)]
as_offset_image = False
if as_offset_image:
offset_img_list = []
artist_list = []
# THIS DOES NOT DO WHAT I WANT
# Scales the image with data coords
from matplotlib.offsetbox import OffsetImage, AnnotationBbox
bboxkw = dict(
xycoords='data',
boxcoords='offset points',
#boxcoords='data',
pad=0.25,
# frameon=False, bboxprops=dict(fc="cyan"),
# arrowprops=dict(arrowstyle="->"))
)
for pos, img, framewidth in zip(pos_list, img_list_, framewidth_list):
offset_img = OffsetImage(img, zoom=.4)
bboxkw['frameon'] = framewidth > 0
artist = AnnotationBbox(offset_img, pos, xybox=(-0., 0.), **bboxkw)
offset_img_list.append(offset_img)
artist_list.append(artist)
else:
# THIS DOES EXACTLY WHAT I WANT
# Ties the image to data coords
#for pos, img, size, color, framewidth in zip(pos_list, img_list_,
# size_list_, color_list,
# framewidth_list):
for pos, img, size in zip(pos_list, img_list_, size_list_):
bbox = vt.bbox_from_center_wh(pos, size)
extent = vt.extent_from_bbox(bbox)
pt.plt.imshow(img, extent=extent)
imgdat = {
#'offset_img_list': offset_img_list,
#'artist_list': artist_list,
}
return imgdat
def show_name_matches(
ibs,
qaid,
name_daid_list,
name_fm_list,
name_fs_list,
name_H1_list,
name_featflag_list,
qreq_=None,
**kwargs
):
"""
Called from chip_match.py
Args:
ibs (IBEISController): wbia controller object
qaid (int): query annotation id
name_daid_list (list):
name_fm_list (list):
name_fs_list (list):
name_H1_list (list):
name_featflag_list (list):
qreq_ (QueryRequest): query request object with hyper-parameters(default = None)
Kwargs:
draw_fmatches, name_rank, fnum, pnum, colorbar_, nonvote_mode,
fastmode, show_matches, fs, fm_norm, lbl1, lbl2, rect, draw_border,
cmap, H1, H2, scale_factor1, scale_factor2, draw_pts, draw_ell,
draw_lines, show_nMatches, all_kpts, in_image, show_query, draw_lbl,
name_annot_scores, score, rawscore, aid2_raw_rank, show_name,
show_nid, show_aid, show_annot_score, show_truth, name_score,
show_name_score, show_name_rank, show_timedelta
CommandLine:
python -m wbia.viz.viz_matches --exec-show_name_matches
python -m wbia.viz.viz_matches --test-show_name_matches --show
Example:
>>> # DISABLE_DOCTEST
>>> from wbia.viz.viz_matches import * # NOQA
>>> from wbia.algo.hots import chip_match
>>> from wbia.algo.hots import name_scoring
>>> import vtool as vt
>>> from wbia.algo.hots import _pipeline_helpers as plh # NOQA
>>> import numpy as np
>>> func = chip_match.ChipMatch.show_single_namematch
>>> sourcecode = ut.get_func_sourcecode(func, stripdef=True, stripret=True,
>>> strip_docstr=True)
>>> setup = ut.regex_replace('viz_matches.show_name_matches', '#', sourcecode)
>>> homog = False
>>> print(ut.indent(setup, '>>> '))
>>> ibs, qreq_, cm_list = plh.testdata_post_sver('PZ_MTEST', qaid_list=[1])
>>> cm = cm_list[0]
>>> cm.score_name_nsum(qreq_)
>>> dnid = ibs.get_annot_nids(cm.qaid)
>>> # +--- COPIED SECTION
>>> locals_ = locals()
>>> var_list = ut.exec_func_src(
>>> func, locals_=locals_,
>>> sentinal='name_annot_scores = cm.annot_score_list.take(sorted_groupxs')
>>> exec(ut.execstr_dict(var_list))
>>> # L___ COPIED SECTION
>>> kwargs = {}
>>> show_name_matches(ibs, qaid, name_daid_list, name_fm_list,
>>> name_fs_list, name_h1_list, name_featflag_list,
>>> qreq_=qreq_, **kwargs)
>>> ut.quit_if_noshow()
>>> ut.show_if_requested()
"""
# logger.info("SHOW NAME MATCHES")
# logger.info(ut.repr2(kwargs, nl=True))
# from wbia import constants as const
from wbia import tag_funcs
draw_fmatches = kwargs.pop('draw_fmatches', True)
rchip1, kpts1 = get_query_annot_pair_info(ibs, qaid, qreq_, draw_fmatches)
rchip2_list, kpts2_list = get_data_annot_pair_info(
ibs, name_daid_list, qreq_, draw_fmatches
)
heatmask = kwargs.pop('heatmask', False)
if heatmask:
from vtool.coverage_kpts import make_kpts_heatmask
import numpy as np
import vtool as vt
wh1 = vt.get_size(rchip1)
fx1 = np.unique(np.hstack([fm.T[0] for fm in name_fm_list]))
heatmask1 = make_kpts_heatmask(kpts1[fx1], wh1)
rchip1 = vt.overlay_alpha_images(heatmask1, rchip1)
# Hack cast back to uint8
rchip1 = (rchip1 * 255).astype(np.uint8)
rchip2_list_ = rchip2_list
rchip2_list = []
for rchip2, kpts2, fm in zip(rchip2_list_, kpts2_list, name_fm_list):
fx2 = fm.T[1]
wh2 = vt.get_size(rchip2)
heatmask2 = make_kpts_heatmask(kpts2[fx2], wh2)
rchip2 = vt.overlay_alpha_images(heatmask2, rchip2)
# Hack cast back to uint8
rchip2 = (rchip2 * 255).astype(np.uint8)
rchip2_list.append(rchip2)
#
fm_list = name_fm_list
fs_list = name_fs_list
featflag_list = name_featflag_list
offset_list, sf_list, bbox_list = show_multichip_match(
rchip1, rchip2_list, kpts1, kpts2_list, fm_list, fs_list, featflag_list, **kwargs
)
aid_list = [qaid] + name_daid_list
annotate_matches3(
ibs,
aid_list,
bbox_list,
offset_list,
name_fm_list,
name_fs_list,
qreq_=None,
**kwargs
)
ax = pt.gca()
title = vh.get_query_text(ibs, None, name_daid_list, False, qaid=qaid, **kwargs)
pt.set_title(title, ax)
# Case tags
annotmatch_rowid_list = ibs.get_annotmatch_rowid_from_superkey(
[qaid] * len(name_daid_list), name_daid_list
)
annotmatch_rowid_list = ut.filter_Nones(annotmatch_rowid_list)
tags_list = ibs.get_annotmatch_case_tags(annotmatch_rowid_list)
if not ut.get_argflag('--show'): # False:
tags_list = tag_funcs.consolodate_annotmatch_tags(tags_list)
tag_list = ut.unique_ordered(ut.flatten(tags_list))
name_rank = kwargs.get('name_rank', None)
truth = get_multitruth(ibs, aid_list)
xlabel = {1: 'Correct ID', 0: 'Incorrect ID', 2: 'Unknown ID'}[truth]
if False:
if name_rank is None:
xlabel = {1: 'Genuine', 0: 'Imposter', 2: 'Unknown'}[truth]
# xlabel = {1: 'True', 0: 'False', 2: 'Unknown'}[truth]
else:
if name_rank == 0:
xlabel = {1: 'True Positive', 0: 'False Positive', 2: 'Unknown'}[truth]
else:
xlabel = {1: 'False Negative', 0: 'True Negative', 2: 'Unknown'}[truth]
if len(tag_list) > 0:
xlabel += '\n' + ', '.join(tag_list)
noshow_truth = ut.get_argflag('--noshow_truth')
if not noshow_truth:
pt.set_xlabel(xlabel)
return ax
def show_sv(chip1, chip2, kpts1, kpts2, fm, homog_tup=None, aff_tup=None,
mx=None, show_assign=True, show_lines=True, show_kpts=True,
show_aff=None, fnum=1, refine_method=None, **kwargs):
""" Visualizes spatial verification
CommandLine:
python -m vtool.spatial_verification --test-spatially_verify_kpts --show
"""
#import plottool as pt
# GEt Matching chips
kpts1_m = kpts1[fm.T[0]]
kpts2_m = kpts2[fm.T[1]]
wh2 = vt.get_size(chip2)
#
# Get Affine Chips, Keypoints, Inliers
if show_aff is None:
show_aff_ = aff_tup is not None
else:
show_aff_ = show_aff
if show_aff_:
(aff_inliers, Aff) = aff_tup
chip1_At = vt.warpAffine(chip1, Aff, wh2)
#kpts1_mAt = ktool.transform_kpts(kpts1_m, Aff)
chip2_blendA = vt.blend_images(chip1_At, chip2)
#
# Get Homog Chips, Keypoints, Inliers
show_homog = homog_tup is not None
if show_homog:
(hom_inliers, H) = homog_tup
#kpts1_mHt = ktool.transform_kpts(kpts1_m, H)
chip1_Ht = vt.warpHomog(chip1, H, wh2)
chip2_blendH = vt.blend_images(chip1_Ht, chip2)
#
# Drawing settings
nRows = (show_assign) + (show_aff_) + (show_homog)
nCols1 = (show_assign) + (show_aff_) + (show_homog)
nCols2 = 2
pnum1_ = df2.get_pnum_func(nRows, nCols1)
pnum2_ = df2.get_pnum_func(nRows, nCols2)
#in_kwargs = dict(rect=True, ell_alpha=.7, eig=False, ori=True, pts=True)
#out_kwargs = dict(rect=False, ell_alpha=.3, eig=False)
in_kwargs = dict(rect=False, ell_alpha=.7, eig=False, ori=False, pts=False)
out_kwargs = dict(rect=False, ell_alpha=.7, eig=False, ori=False)
def _draw_kpts(*args, **kwargs):
if not show_kpts:
return
df2.draw_kpts2(*args, **kwargs)
def draw_inlier_kpts(kpts_m, inliers, color, H=None):
_draw_kpts(kpts_m[inliers], color=color, H=H, **in_kwargs)
if mx is not None:
_draw_kpts(kpts_m[mx:(mx + 1)], color=color, ell_linewidth=3, H=H, **in_kwargs)
def _draw_matches(px, title, inliers):
dmkwargs = dict(fs=None, title=title, all_kpts=False, draw_lines=True,
docla=True, draw_border=True, fnum=fnum, pnum=pnum1_(px), colors=df2.ORANGE)
__fm = np.vstack((inliers, inliers)).T
df2.show_chipmatch2(chip1, chip2, kpts1_m, kpts2_m, __fm, **dmkwargs)
return px + 1
from plottool import color_funcs
colors = df2.distinct_colors(2, brightness=.95)
color1, color2 = colors[0], colors[1]
color1_dark = color_funcs.darken_rgb(color1, .2)
color2_dark = color_funcs.darken_rgb(color2, .2)
def _draw_chip(px, title, chip, inliers, kpts1_m, kpts2_m, H1=None):
if isinstance(px, tuple):
pnum = px
df2.imshow(chip, title=title, fnum=fnum, pnum=pnum)
px = pnum[2]
else:
df2.imshow(chip, title=title, fnum=fnum, pnum=pnum2_(px))
if kpts1_m is not None:
_draw_kpts(kpts1_m, color=color1_dark, H=H1, **out_kwargs)
draw_inlier_kpts(kpts1_m, inliers, color1, H=H1)
if kpts2_m is not None:
_draw_kpts(kpts2_m, color=color2_dark, **out_kwargs)
draw_inlier_kpts(kpts2_m, inliers, color2)
if kpts2_m is not None and kpts1_m is not None and show_lines:
__fm = np.vstack((inliers, inliers)).T
df2.draw_lines2(kpts1_m, kpts2_m, __fm,
color_list=[custom_constants.ORANGE], lw=2,
line_alpha=1,
H1=H1)
return px + 1
#
# Begin the drawing
df2.figure(fnum=fnum, pnum=(nRows, nCols1, 1), docla=True, doclf=True)
px = 0
if show_assign:
# Draw the Assigned -> Affine -> Homography matches
px = _draw_matches(px, '%d Assigned matches ' % len(fm), np.arange(len(fm)))
if show_aff_:
px = _draw_matches(px, '%d Initial inliers ' % len(aff_inliers), aff_inliers)
if show_homog:
if refine_method is None:
refine_method = ''
if len(refine_method) > 0:
refine_method_ = '(%s) ' % (refine_method,)
else:
refine_method_ = ''
px = _draw_matches(
px,
'%d Refined %sinliers' % (len(hom_inliers), refine_method_),
hom_inliers)
#
# Draw the Affine Transformations
px = nCols2 * show_assign
#if show_aff_ or show_homog:
if show_aff_:
#px = _draw_chip(px, 'Source', chip1, aff_inliers, kpts1_m, None)
#px = _draw_chip(px, 'Dest', chip2, aff_inliers, None, kpts2_m)
px = _draw_chip(px, 'Initial Warped', chip1_At, aff_inliers, kpts1_m, None, H1=Aff)
px = _draw_chip(px, 'Initial Blend', chip2_blendA, aff_inliers, kpts1_m, kpts2_m, H1=Aff)
#px = _draw_chip(px, 'Affine', chip1_At, aff_inliers, kpts1_mAt, None)
#px = _draw_chip(px, 'Aff Blend', chip2_blendA, aff_inliers, kpts1_mAt, kpts2_m)
pass
#
# Draw the Homography Transformation
if show_homog:
#px = _draw_chip(px, 'Source', chip1, hom_inliers, kpts1_m, None)
#px = _draw_chip(px, 'Dest', chip2, hom_inliers, None, kpts2_m)
#px = _draw_chip(px, 'Homog', chip1_Ht, hom_inliers, kpts1_mHt, None)
#px = _draw_chip(px, 'Homog Blend', chip2_blendH, hom_inliers, kpts1_mHt, kpts2_m)
px = _draw_chip(px, 'Refined Warped', chip1_Ht, hom_inliers, kpts1_m, None, H1=H)
px = _draw_chip(px, 'Refined Blend', chip2_blendH, hom_inliers, kpts1_m, kpts2_m, H1=H)
