diff --git a/.gitignore b/.gitignore
index 221a58901f03894b128b24ad8afbe23032c9fd45..a3b8a68577ea77305d2462b4e8bd46bbb702c7ef 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,3 +1,11 @@
+env/
testing/output.mp4
testing/abc.png
+*.o
+*.so
+*.pyc
+*.egg-info
+fastXPCS/fastXPCS/cython/welford.c
+fastXPCS/fastXPCS/cython/fxpcs_sparse.c
+fastXPCS/fastXPCS/cython/dense.c
diff --git a/fastXPCS/__init__.py b/fastXPCS/fastXPCS/__init__.py
similarity index 100%
rename from fastXPCS/__init__.py
rename to fastXPCS/fastXPCS/__init__.py
diff --git a/fastXPCS/algos.py b/fastXPCS/fastXPCS/algos.py
similarity index 100%
rename from fastXPCS/algos.py
rename to fastXPCS/fastXPCS/algos.py
diff --git a/fastXPCS/processOnCPU.py b/fastXPCS/fastXPCS/processOnCPU.py
similarity index 100%
rename from fastXPCS/processOnCPU.py
rename to fastXPCS/fastXPCS/processOnCPU.py
diff --git a/install_env.sh b/install_env.sh
new file mode 100644
index 0000000000000000000000000000000000000000..d84734af8a95bbe8307d40271c9f140316cf9a6d
--- /dev/null
+++ b/install_env.sh
@@ -0,0 +1,23 @@
+
+python3 -m pip install --upgrade pip
+
+python3 -m venv env
+source env/bin/activate
+
+python3 -m pip install Cython
+python3 -m pip install numpy
+python3 -m pip install scipy
+python3 -m pip install matplotlib
+python3 -m pip install pillow
+
+cd fastXPCS
+
+#python setup.py build_ext --inplace
+#python setup.py develop
+
+python3 -m pip install -e .
+cd ..
+
+
+
+
diff --git a/run_test.sh b/run_test.sh
new file mode 100644
index 0000000000000000000000000000000000000000..5086dc94104014252128458c37a7f6073bcfa09e
--- /dev/null
+++ b/run_test.sh
@@ -0,0 +1,3 @@
+source env/bin/activate
+
+python3 testing/test.py
\ No newline at end of file
diff --git a/testing/test.py b/testing/test.py
index ecc4de0e430a7b253eb473d6fa91fefae8749fa9..a111fdd9bcea9d679236510ad8e5fda1ab05fd7a 100644
--- a/testing/test.py
+++ b/testing/test.py
@@ -6,6 +6,9 @@ import numpy as np
import matplotlib.pyplot as pl
# get some data
+from fastXPCS.algos import TTCdata, do_sparse_train
+from fastXPCS.fxpcs_sparse import sparsify, sparsify2
+from PIL import Image
dataShape = (352, 512, 128)
photonEnergy = 13.3 # in ADU
@@ -44,6 +47,7 @@ def getDummyData(shape, chanceOfPhotonPerPixel = 0.01, correlationPerBunch=0.5,
p=[q**0, q**1, q**2, q**3, q**4, q**5, q**6]/np.sum([q**0, q**1, q**2, q**3, q**4, q**5, q**6])),(shape[1], shape[2]))
def evolveImageFrom(image, c=correlationPerBunch):
+ print(c)
randMask = np.reshape(np.random.choice(
[0, 1],
size=numPixelPrImage,
@@ -54,7 +58,8 @@ def getDummyData(shape, chanceOfPhotonPerPixel = 0.01, correlationPerBunch=0.5,
data[0,:,:] = np.reshape(getRandomImage(), (shape[1], shape[2]))
for pulseNr in range(1,dataShape[0]):
- data[pulseNr, :, :] = evolveImageFrom(data[pulseNr-1,:,:])
+ correlationPerBunch += (1-correlationPerBunch)*0.001
+ data[pulseNr, :, :] = evolveImageFrom(data[pulseNr-1,:,:], correlationPerBunch )
# if we do not returned photnoized data, then add a gaussian random variable
if (~photonized ):
@@ -70,15 +75,26 @@ data = getDummyData(
correlationPerBunch=0.5,
photonized=False,
energy=photonEnergy,
- sigma=0.12)
+ sigma=0.012)
# function to do the xpcs the same way as in the xpcs class
# in the xpcs calng addon
# input : -image, the data set, e.g., data
def xpcs(image, energy):
- image /= 1
+
+
+ _ttcdata = TTCdata()
+ _ttcdata.user_q_mask = np.ones(shape=(1,512,128)).astype(np.uint32)
+ _ttcdata.current_cal_mask = np.ones(shape=(1,512,128)).astype(np.uint32)
+ print(_ttcdata.user_q_mask.shape)
+ print(_ttcdata.current_cal_mask.shape)
+ _sparse_array = None
+ _cal_mask = None
+
+ image /= energy
image = np.around(image)
+ print(f" {np.sum(image==0)} num zeros {image.size}")
flat_image_shape = image.reshape(image.shape[0], -1).shape
print(flat_image_shape)
@@ -90,24 +106,59 @@ def xpcs(image, energy):
# lit, i.e., have a value larger than 0
lit_mask = reshaped_data > 0
# get the values of those that have a value larger than 0
- lit_pixels = reshaped_data[lit_mask].astype(np.uint32)
+ lit_pixels = np.ascontiguousarray(reshaped_data[lit_mask].astype(np.uint32))
# get the indices of those that
- lit_indices = indices[:, lit_mask].astype(np.uint32)
+ lit_indices = np.ascontiguousarray(indices[:, lit_mask].astype(np.uint32))
# initalize a sparse array
sparse_array = np.empty_like(lit_pixels, dtype=np.uint32)
- # call function to fill sparse array with indices and value via bitpacking
+ # call function to fill sparse array with indices and value via bitpacking (from Felix)
+ # note that aliaksandr.leonau@xfel.eu mentioned that using build in sparse capabilities from
+ # numpy/scipy gives comparable or better results.
sparsify2(lit_pixels, lit_indices, sparse_array)
+ image = image.astype(np.int32)
+ ## sparsifing again?
+
+ start = time.time()
+ sparse_ref = sparsify(image.reshape(image.shape[0], -1))
+
+ _ttcdata.current_image = image.astype(np.int32)
+ _ttcdata.current_image_sparse = sparse_ref
+ _ttcdata.current_cal_mask = np.isnan(image)
+
+ do_sparse_train(_ttcdata)
+ elapsed = time.time() - start
+ print(f"TTCF in: {elapsed:.3f}")
+
+ output = _ttcdata.ttc_on.astype(np.int32)
+ return output
+
# do stuff
t = time.time()
-xpcs(data, photonEnergy)
+out = xpcs(data, photonEnergy)
elapsed = time.time() - t
print(f"elapsed {elapsed}")
-
+print(out.shape)
+print(np.min(out))
+print(np.max(out))
+def plot(data):
+ data = data.astype(np.float32)
+ cmap = pl.get_cmap('viridis')
+ norm = pl.Normalize(data.min(), data.max())
+ print("blib")
+ # The norm instance scales data to a 0-1 range, cmap makes it RGB
+ rgb = cmap(norm(data))
+ print("blub")
+ # MPL uses a 0-1 float RGB representation, so we'll scale to 0-255
+ rgb = (255 * rgb).astype(np.uint8)
+ print("blab")
+ Image.fromarray(rgb).save('test.png')
+
+plot(out[0,:,:])
# cpu code with python from xpcs