# -*- coding: utf-8 -*-
# Copyright (c) 2020 Stephen Wasilewski, HSLU and EPFL
# =======================================================================
# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
# =======================================================================
import numpy as np
from raytraverse import translate
from raytraverse.lightfield.lightplanekd import LightPlaneKD
[docs]class SensorPlaneKD(LightPlaneKD):
"""collection of sensor results with KDtree structure for positional
query
data has shape (pts, sensors, sources, bands)
"""
@property
def sensors(self):
return self._sensors
@property
def vecs(self):
"""indexing vectors (such as position, sun positions, etc.)"""
return self._vecs
@vecs.setter
def vecs(self, pt):
if pt is None:
pt = f"{self.scene.outdir}/{self.pm.name}/{self.src}.npz"
with np.load(pt) as result:
self._vecs = result['vecs']
self._sensors = result['sensors']
self.data = result['lum']
self._kd = None
if self._vecs.shape[-1] > 3:
self._samplelevel = self.vecs[:, 0]
self._vecs = self._vecs[:, -3:]
else:
self._samplelevel = np.zeros(len(self.vecs))
self.omega = None
@property
def data(self):
"""light data"""
return self._data
@data.setter
def data(self, lum):
self._data = lum
[docs] @staticmethod
def apply_coef(data, coefs):
"""apply coefficient vector to data
Parameters
----------
data: np.array
ndims should match self.data (N, sensors, nsrcs, nfeatures)
coefs: np.array int float list
shape (L, self.srcn) or broadcastable
Returns
-------
alum: np.array
shape (L, N, sensors, nfeatures)
"""
features = data.shape[-1]
srcn = data.shape[-2]
if features > 1:
try:
c = np.asarray(coefs).reshape(-1, srcn, features)
except ValueError:
c = np.broadcast_to(coefs, (1, srcn, features))
sstring = 'abc,debc->adec'
else:
try:
c = np.asarray(coefs).reshape(-1, srcn)
except ValueError:
c = np.broadcast_to(coefs, (1, srcn))
sstring = 'ab,debc->adec'
return np.einsum(sstring, c, data)
[docs] def evaluate(self, skyvec, points=None, sensoridx=None, mask=True,
**kwargs):
if points is None:
qidx = np.arange(len(self.vecs))
else:
qidx, d = self.query(points)
if mask:
omask = self.pm.mask
self.pm.mask = True
qidx = qidx[self.pm.in_view(points, False)]
self.pm.mask = omask
if len(qidx) > len(self.data):
data = translate.simple_take(self.data, sensoridx, axes=1)
r = self.apply_coef(data, skyvec)[:, qidx]
else:
data = translate.simple_take(self.data, qidx, sensoridx)
r = self.apply_coef(data, skyvec)
return r
[docs] def direct_view(self, res=512, showsample=True, area=False,
interp=False, sensoridx=None, **kwargs):
"""create a summary image of lightplane showing samples and areas"""
if area:
outf = self._datadir.replace("/", "_") + f"{self.src}_area.hdr"
self.make_image(outf, self.omega, res=res, showsample=showsample,
interp=False)
result = self.evaluate(1, sensoridx=sensoridx)
if sensoridx is None:
sensoridx = np.arange(result.shape[-2])
else:
sensoridx = np.ravel(sensoridx)
for i, j in enumerate(sensoridx):
vd = "vd-{:.1f}_{:.1f}_{:.1f}".format(*self.sensors[j, 3:])
outf = self._datadir.replace("/", "_") + f"{self.src}_{vd}.hdr"
self.make_image(outf, result[0, :, i, 0], res=res,
showsample=showsample, interp=interp)
