foscat 2025.11.1__tar.gz → 2026.2.1__tar.gz

{foscat-2025.11.1/src/foscat.egg-info → foscat-2026.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: foscat
-Version: 2025.11.1
+Version: 2026.2.1
 Summary: Generate synthetic Healpix or 2D data using Cross Scattering Transform
 Author-email: Jean-Marc DELOUIS <jean.marc.delouis@ifremer.fr>
 Maintainer-email: Theo Foulquier <theo.foulquier@ifremer.fr>
@@ -91,74 +91,6 @@ git clone https://github.com/jmdelouis/FOSCAT_DEMO.git
 
 ```
 
-# Spherical data example
-
-## compute a synthetic image
-
-```
-python demo.py -n=32 -k -c -s=100
-```
-
-The _demo.py_ script serves as a demonstration of the capabilities of the foscat library. It utilizes the Cross Wavelet Scattering Transform to generate a Healpix map that possesses the same characteristics as a specified input map.
-
-- `-n=32` computes map with nside=32.
-- `-k` uses 5x5 kernel.
-- `-c` uses Scattering Covariance.
-- `-l` uses LBFGS minimizer.
-- `-s=100` computes 100 steps.
-
-```
-python demo.py -n=8 [-c|--cov][-s|--steps=3000][-S=1234|--seed=1234][-k|--k5x5][-d|--data][-o|--out][-r|--orient] [-p|--path][-a|--adam]
-
-```
-
-- The "-n" option specifies the nside of the input map. The maximum nside value is 256 with the default map.
-- The "--cov" option (optional) uses scat_cov instead of scat.
-- The "--steps" option (optional) specifies the number of iterations. If not specified, the default value is 1000.
-- The "--seed" option (optional) specifies the seed of the random generator.
-- The "--path" option (optional) allows you to define the path where the output files will be written. The default path is "data".
-- The "--k5x5" option (optional) uses a 5x5 kernel instead of a 3x3.
-- The "--data" option (optional) specifies the input data file to be used. If not specified, the default file "LSS_map_nside128.npy" will be used.
-- The "--out" option (optional) specifies the output file name. If not specified, the output file will be saved in "demo".
-- The "--orient" option (optional) specifies the number of orientations. If not specified, the default value is 4.
-- The "--adam" option (optional) makes the synthesis using the ADAM optimizer instead of the L_BFGS.
-
-## plot the result
-
-The following script generates a series of plots that showcase different aspects of the synthesis process using the _demo.py_ script.
-
-> python test2D.py
-
-```
-python plotdemo.py -n=32 -c
-```
-
-# 2D field demo
-
-> python test2Dplot.py
-
-# compute a synthetic turbulent field
-
-The python scripts _demo2D.py_ included in this package demonstrate how to use the foscat library to generate a 2D synthetic fields that have patterns with the same statistical properties as a specified 2D image. In this particular case, the input field is a sea surface temperature extracted from a north atlantic ocean simulation.
-
-> python testHealpix.py
-
-```
-python demo2d.py -n=32 -k -c
-```
-
-> python testHplot.py
-
-The following script generates a series of plots that showcase different aspects of the synthesis process using the _demo2D.py_ script.
-
-```
-python plotdemo2d.py -n=32 -c
-```
-
-For more information, see the [documentation](https://foscat-documentation.readthedocs.io/en/latest/index.html).
-
-> mpirun -np 3 testHealpix_mpi.py
-
 ## Authors and acknowledgment
 
 Authors: J.-M. Delouis, P. Campeti, T. Foulquier, J. Mangin, L. Mousset, T. Odaka, F. Paul, E. Allys

{foscat-2025.11.1 → foscat-2026.2.1}/README.md

@@ -62,74 +62,6 @@ git clone https://github.com/jmdelouis/FOSCAT_DEMO.git
 
 ```
 
-# Spherical data example
-
-## compute a synthetic image
-
-```
-python demo.py -n=32 -k -c -s=100
-```
-
-The _demo.py_ script serves as a demonstration of the capabilities of the foscat library. It utilizes the Cross Wavelet Scattering Transform to generate a Healpix map that possesses the same characteristics as a specified input map.
-
-- `-n=32` computes map with nside=32.
-- `-k` uses 5x5 kernel.
-- `-c` uses Scattering Covariance.
-- `-l` uses LBFGS minimizer.
-- `-s=100` computes 100 steps.
-
-```
-python demo.py -n=8 [-c|--cov][-s|--steps=3000][-S=1234|--seed=1234][-k|--k5x5][-d|--data][-o|--out][-r|--orient] [-p|--path][-a|--adam]
-
-```
-
-- The "-n" option specifies the nside of the input map. The maximum nside value is 256 with the default map.
-- The "--cov" option (optional) uses scat_cov instead of scat.
-- The "--steps" option (optional) specifies the number of iterations. If not specified, the default value is 1000.
-- The "--seed" option (optional) specifies the seed of the random generator.
-- The "--path" option (optional) allows you to define the path where the output files will be written. The default path is "data".
-- The "--k5x5" option (optional) uses a 5x5 kernel instead of a 3x3.
-- The "--data" option (optional) specifies the input data file to be used. If not specified, the default file "LSS_map_nside128.npy" will be used.
-- The "--out" option (optional) specifies the output file name. If not specified, the output file will be saved in "demo".
-- The "--orient" option (optional) specifies the number of orientations. If not specified, the default value is 4.
-- The "--adam" option (optional) makes the synthesis using the ADAM optimizer instead of the L_BFGS.
-
-## plot the result
-
-The following script generates a series of plots that showcase different aspects of the synthesis process using the _demo.py_ script.
-
-> python test2D.py
-
-```
-python plotdemo.py -n=32 -c
-```
-
-# 2D field demo
-
-> python test2Dplot.py
-
-# compute a synthetic turbulent field
-
-The python scripts _demo2D.py_ included in this package demonstrate how to use the foscat library to generate a 2D synthetic fields that have patterns with the same statistical properties as a specified 2D image. In this particular case, the input field is a sea surface temperature extracted from a north atlantic ocean simulation.
-
-> python testHealpix.py
-
-```
-python demo2d.py -n=32 -k -c
-```
-
-> python testHplot.py
-
-The following script generates a series of plots that showcase different aspects of the synthesis process using the _demo2D.py_ script.
-
-```
-python plotdemo2d.py -n=32 -c
-```
-
-For more information, see the [documentation](https://foscat-documentation.readthedocs.io/en/latest/index.html).
-
-> mpirun -np 3 testHealpix_mpi.py
-
 ## Authors and acknowledgment
 
 Authors: J.-M. Delouis, P. Campeti, T. Foulquier, J. Mangin, L. Mousset, T. Odaka, F. Paul, E. Allys

{foscat-2025.11.1 → foscat-2026.2.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "foscat"
-version = "2025.11.1"
+version = "2026.02.1"
 description = "Generate synthetic Healpix or 2D data using Cross Scattering Transform"
 readme = "README.md"
 license = { text = "BSD-3-Clause" }

{foscat-2025.11.1 → foscat-2026.2.1}/src/foscat/FoCUS.py

@@ -5,8 +5,11 @@ import healpy as hp
 import numpy as np
 import foscat.HealSpline as HS
 from scipy.interpolate import griddata
+from foscat.SphereDownGeo import SphereDownGeo
+from foscat.SphereUpGeo import SphereUpGeo
+import torch
 
-TMPFILE_VERSION = "V10_0"
+TMPFILE_VERSION = "V13_0"
 
 
 class FoCUS:
@@ -36,7 +39,7 @@ class FoCUS:
             mpi_rank=0
     ):
 
-        self.__version__ = "2025.11.1"
+        self.__version__ = "2026.02.1"
         # P00 coeff for normalization for scat_cov
         self.TMPFILE_VERSION = TMPFILE_VERSION
         self.P1_dic = None
@@ -57,7 +60,8 @@ class FoCUS:
         self.kernelR_conv = {}
         self.kernelI_conv = {}
         self.padding_conv = {}
-
+        self.down = {}
+        self.up = {}
         if not self.silent:
             print("================================================")
             print("          START FOSCAT CONFIGURATION")
@@ -648,6 +652,7 @@ class FoCUS:
         return rim
 
     # --------------------------------------------------------
+    
     def ud_grade_2(self, im, axis=0, cell_ids=None, nside=None,max_poll=False):
 
         if self.use_2D:
@@ -721,6 +726,22 @@ class FoCUS:
 
         else:
             shape = list(im.shape)
+            if nside is None:
+                l_nside=int(np.sqrt(shape[-1]//12))
+            else:
+                l_nside=nside
+            
+            nbatch=1
+            for k in range(len(shape)-1):
+                nbatch*=shape[k]
+            if l_nside not in self.down:
+                print('initialise down', l_nside)
+                self.down[l_nside] = SphereDownGeo(nside_in=l_nside, dtype=self.all_bk_type,mode="smooth", in_cell_ids=cell_ids)
+            
+            res,out_cell=self.down[l_nside](self.backend.bk_reshape(im,[nbatch,1,shape[-1]]))
+            
+            return self.backend.bk_reshape(res,shape[:-1]+[out_cell.shape[0]]),out_cell
+            '''
             if self.use_median:
                 if cell_ids is not None:
                     sim, new_cell_ids = self.backend.binned_mean(im, cell_ids,reduce='median')
@@ -747,6 +768,7 @@ class FoCUS:
                 return self.backend.bk_reduce_mean(
                     self.backend.bk_reshape(im, shape[0:-1]+[shape[-1]//4,4]), axis=-1
                 ),None
+            '''
 
     # --------------------------------------------------------
     def up_grade(self, im, nout,
@@ -836,6 +858,7 @@ class FoCUS:
             else:
                 lout = nside
                 
+            '''
             if (lout,nout) not in self.pix_interp_val or force_init_index:
                 if not self.silent:
                     print("compute lout nout", lout, nout)
@@ -926,12 +949,32 @@ class FoCUS:
                     
                     del w
                     del p
-
-            if lout == nout:
-                imout = im
-            else:
-                # work only on the last column
-
+            '''
+            shape=list(im.shape)
+            nbatch=1
+            for k in range(len(shape)-1):
+                nbatch*=shape[k]
+            
+            im=self.backend.bk_reshape(im,[nbatch,1,shape[-1]])
+            
+            while lout<nout:
+                if lout not in self.up:
+                    if o_cell_ids is None:
+                        l_o_cell_ids=torch.tensor(np.arange(12*(lout**2),dtype='int'),device=im.device)
+                    else:
+                        l_o_cell_ids=o_cell_ids
+                    self.up[lout] = SphereUpGeo(nside_out=lout, 
+                                                dtype=self.all_bk_type,
+                                                 cell_ids_out=l_o_cell_ids,
+                                                 up_norm="col_l1")
+                im, fine_ids = self.up[lout](self.backend.bk_cast(im))
+                lout*=2
+                if lout<nout and o_cell_ids is not None:
+                    o_cell_ids=torch.repeat(fine_ids,4)*4+ \
+                                torch.tile(torch.tensor([0,1,2,3],device=fine_ids.device,dtype=fine_ids.dtype),fine_ids.shape[0])
+            
+            return self.backend.bk_reshape(im,shape[:-1]+[im.shape[-1]])
+            '''
                 ndata = 1
                 for k in range(len(ishape)-1):
                     ndata = ndata * ishape[k]
@@ -960,6 +1003,7 @@ class FoCUS:
                     return self.backend.bk_reshape(
                         imout, ishape[0:-1]+[imout.shape[-1]]
                     )
+            '''
         return imout
 
     # --------------------------------------------------------
@@ -1354,7 +1398,9 @@ class FoCUS:
                     else:
                         l_cell_ids=cell_ids
                         
-                    nvalid=self.KERNELSZ**2
+                    nvalid=4*self.KERNELSZ**2
+                    if nvalid>12*nside**2:
+                        nvalid=12*nside**2
                     idxEB=hconvol.idx_nn[:,0:nvalid]
                     tmpEB=np.zeros([self.NORIENT,4,l_cell_ids.shape[0],nvalid],dtype='complex')       
                     tmpS=np.zeros([4,l_cell_ids.shape[0],nvalid],dtype='float')           
@@ -1500,7 +1546,7 @@ class FoCUS:
                     
                 else:
                     if l_kernel == 5:
-                        pw = 0.5
+                        pw = 0.75
                         pw2 = 0.5
                         threshold = 2e-5
 
@@ -1533,11 +1579,20 @@ class FoCUS:
                     
                     xx=np.tile(np.arange(self.KERNELSZ)-self.KERNELSZ//2,self.KERNELSZ).reshape(self.KERNELSZ,self.KERNELSZ)
                     
-                    wwr=(np.exp(-pw2*(xx**2+(xx.T)**2))*np.cos(pw*xx*np.pi)).reshape(1,1,self.KERNELSZ*self.KERNELSZ)
-                    wwr-=wwr.mean()
-                    wwi=(np.exp(-pw2*(xx**2+(xx.T)**2))*np.sin(pw*xx*np.pi)).reshape(1,1,self.KERNELSZ*self.KERNELSZ)
-                    wwi-=wwi.mean()
-                    amp=np.sum(abs(wwr+1J*wwi))
+                    if nside>2:
+                        wwr=(np.exp(-pw2*(xx**2+(xx.T)**2))*np.cos(pw*xx*np.pi)).reshape(1,1,self.KERNELSZ*self.KERNELSZ)
+                        wwr-=wwr.mean()
+                        wwi=(np.exp(-pw2*(xx**2+(xx.T)**2))*np.sin(pw*xx*np.pi)).reshape(1,1,self.KERNELSZ*self.KERNELSZ)
+                        wwi-=wwi.mean()
+                        amp=np.sum(abs(wwr+1J*wwi))
+                    else:
+                        #asymetric kernels
+                        wwr=(np.exp(-2*pw2*(xx**2+self.NORIENT*(xx.T)**2))).reshape(1,1,self.KERNELSZ*self.KERNELSZ)
+                        #wwr-=wwr.mean()
+                        #wwi=(np.exp(-pw2*(xx**2+(xx.T)**2))*np.sin(pw*xx*np.pi)).reshape(1,1,self.KERNELSZ*self.KERNELSZ)
+                        #wwi-=wwi.mean()
+                        wwi=0.0*wwr
+                        amp=self.NORIENT*np.sum(abs(wwr+1J*wwi))
                     
                     wwr/=amp
                     wwi/=amp