def __read_datalist(self):
""" Read json filelist """
# - Check datalist files
if self.datalistfile == "" or self.datalistfile_mask == "":
logger.error("Data list files are empty!")
return -1
# - Read data list for images and store number of instances per class, etc
ret = self.__read_filelist(self.datalistfile)
if ret is None:
logger.error("Failed to read filelist for imgs!")
return -1
datadict = ret[0]
nchannels_set = ret[1]
self.datalist = datadict["data"]
self.nchannels = list(nchannels_set)[0]
self.datasize = len(self.datalist)
self.labels = [item["label"] for item in self.datalist]
self.snames = [item["sname"] for item in self.datalist]
self.classids = [item["id"] for item in self.datalist]
self.classfract_map = dict(Counter(self.classids).items())
logger.info("#%d objects in dataset" % self.datasize)
return 0
def __run_aereco(self):
""" Run AE reconstruction """
# - Set FeatExtractorAE class
ae = FeatExtractorAE(self.dl)
ae.set_image_size(self.nx, self.ny)
ae.normalize = self.normalize
ae.scale_to_abs_max = self.scale_to_abs_max
ae.scale_to_max = self.scale_to_max
ae.log_transform_img = self.log_transform
ae.scale_img = self.scale
ae.scale_img_factors = self.scale_factors
ae.standardize_img = self.standardize
ae.img_means = self.img_means
ae.img_sigmas = self.img_sigmas
ae.chan_divide = self.chan_divide
ae.chan_mins = self.chan_mins
ae.erode = self.erode
ae.erode_kernel = self.erode_kernel
ae.add_channorm_layer = self.add_channorm_layer
# - Run AE reco
status = ae.reconstruct_data(self.modelfile_encoder,
self.weightfile_encoder,
self.modelfile_decoder,
self.weightfile_decoder,
winsize=self.winsize,
outfile_metrics=self.outfile,
save_imgs=self.save_imgs)
if status < 0:
logger.error("AE reconstruction failed (see logs)!")
return -1
return 0
def __load_model(self, modelfile, weightfile):
""" Load model and weights from input h5 file """
#==============================
#== LOAD MODEL ARCHITECTURE
#==============================
# - Load model
try:
#self.model = model_from_json(open(modelfile_json).read())
self.model = load_model(modelfile, custom_objects={'recall_metric': recall_metric, 'precision_metric': precision_metric, 'f1score_metric': f1score_metric})
self.model.load_weights(weightfile)
except Exception as e:
logger.warn("Failed to load model from file %s (err=%s)!" % (modelfile, str(e)))
return -1
if not self.model or self.model is None:
logger.error("Model object is None, loading failed!")
return -1
#===========================
#== SET LOSS & METRICS
#===========================
self.model.compile(optimizer=self.optimizer, loss=self.loss_type, metrics=['accuracy', f1score_metric, precision_metric, recall_metric], run_eagerly=True)
return 0
def compute_bkg(self, masks, sigma_clip=3):
""" Compute image background """
# - Init bkg
self.bkg_levels= [0]*self.nchannels
self.bkg_rms= [0]*self.nchannels
if len(masks)!=self.nchannels:
logger.error("Number of input masks != nchannels, cannot compute bkg!")
return -1
# - Compute bkg levels & rms
logger.info("Computing image clipped stats of non-masked pixels ...")
for i in range(self.nchannels):
data= self.img_data[i]
mask= masks[i]
cond= np.logical_and(np.logical_and(data!=0, np.isfinite(data)), mask==0)
data_1d= data[cond]
print("--> data_1d.shape")
print(data_1d.shape)
mean, median, stddev= sigma_clipped_stats(data_1d, sigma=sigma_clip)
self.bkg_levels[i]= median
self.bkg_rms[i]= stddev
return 0
def compute_img_moments(self):
""" Compute image moments """
# - Init pars
self.moments_c= []
self.moments_hu= []
self.moments_zern= []
# - Compute raw moments and centroids
if self.__compute_contour_pars()<0:
logger.error("Failed to compute contour pars!")
return -1
# - Compute moments (central, Hu, Zernike) of intensity images
# NB: use centroid from refch for all channels
centroid= self.centroids[self.refch]
#centroid= self.center_of_masses[self.refch]
for i in range(self.nchannels):
data= self.img_data[i]
radius= self.radii[i]
ret= self.__compute_moments(data, centroid, radius)
if ret is None:
logger.error("Failed to compute moments for image %s (id=%s, ch=%d)!" % (self.sname, self.label, i+1))
return -1
self.moments_c.append(ret[0])
self.moments_hu.append(ret[1])
self.moments_zern.append(ret[2])
return 0
def run(self, datalist):
""" Run data checker """
# - Init
self.param_dict_list = []
# - Read data
logger.info("Read data list %s ..." % (self.datalist))
self.__read_data(datalist)
# - Run AE reco
logger.info("Running autoencoder reconstruction ...")
if self.__run_aereco() < 0:
logger.error("AE reconstruction failed!")
return -1
# - Select AE reco data
logger.info("Reading and thresholding AE reco metrics ...")
if self.__fill_metric_data() < 0:
logger.error("Failed to read and threshold AE reco metrics!")
return -1
# - Save output data
logger.info("Saving output to file ...")
if self.__save() < 0:
logger.warn("Failed to save output data to file %s!" %
(self.outfile))
return -1
return 0
def __select_cols(self, selcols):
""" Select data columns provided in selcols list """
# - Check sel cols
if not selcols:
logger.error("Empty sel col list!")
return -1
# - Remove any duplicated col ids, sort and set colsel flags
selcols = list(set(selcols))
selcols.sort()
selcolflags = [False] * self.nfeatures
for col in selcols:
if col < 0 or col >= self.nfeatures:
logger.error(
"Given sel col id %d is not in nfeature col range [0,%d]!"
% (col, self.nfeatures - 1))
return -1
selcolflags[col] = True
print("--> Selected columns")
print(selcols)
print("--> Selected column flags")
print(selcolflags)
# - Extract selected data columns
logger.info(
"Extracting selected data columns (N=%d) from original data ..." %
(len(selcols)))
self.data_sel = self.data[:, selcolflags]
self.data_preclassified_sel = self.data_preclassified[:, selcolflags]
self.selfeatids = selcols
return 0
def select_features(self, selcolids):
""" Select feature cols (0 index is the first feature, not sname) """
# - Check if param dict is filled
if not self.param_dict or self.param_dict is None:
logger.error("Parameter dict is empty!")
return -1
# - Get list of sel keys given col indices
keys= list(self.param_dict.keys())
keys_sel= [keys[selcol+1] for selcol in selcolids] # +1 because 0 index is the first feature, not sname
# - Create new dict with selected pars
param_dict_sel= collections.OrderedDict()
param_dict_sel["sname"]= self.param_dict["sname"]
for key in keys_sel:
param_dict_sel[key]= self.param_dict[key]
param_dict_sel["id"]= self.param_dict["id"]
# - Override old dict
self.param_dict= param_dict_sel
return 0
def __read_fits(self, filename): """ Read FITS image and return data """ # - Open file try: hdu= fits.open(filename,memmap=False) except Exception as ex: errmsg= 'Cannot read image file: ' + filename #cls._logger.error(errmsg) logger.error(errmsg) raise IOError(errmsg) # - Read data data= hdu[0].data data_size= np.shape(data) nchan= len(data.shape) if nchan==4: output_data= data[0,0,:,:] elif nchan==2: output_data= data else: errmsg= 'Invalid/unsupported number of channels found in file ' + filename + ' (nchan=' + str(nchan) + ')!' #cls._logger.error(errmsg) logger.error(errmsg) hdu.close() raise IOError(errmsg) # - Read metadata header= hdu[0].header # - Close file hdu.close() return output_data, header
def read_img(self):
""" Read input image and generate Montage metadata """
# - Read FITS (ALL PROC)
logger.info("[PROC %d] Reading input image %s ..." % (procId, self.imgfile_fullpath))
try:
data, header, wcs= Utils.read_fits(self.imgfile_fullpath)
except Exception as e:
logger.error("[PROC %d] Failed to read input image %s (err=%s)!" % (procId, self.imgfile_fullpath, str(e)))
return -1
data= ret[0]
header= ret[1]
wcs= ret[2]
# - Write input image Montage metadata (PROC 0)
status= 0
if procId==MASTER:
status= Utils.write_montage_fits_metadata(inputfile=self.imgfile_fullpath, metadata_file=self.img_metadata, jobdir=self.jobdir_scutout)
else: # OTHER PROCS
status= -1
if comm is not None:
status= comm.bcast(status, root=MASTER)
if status<0:
logger.error("[PROC %d] Failed to generate Montage metadata for input image %s, exit!" % (procId, self.imgfile_fullpath))
return -1
return 0
def __load_model(self, modelfile):
""" Load model and weights from input h5 file """
#==============================
#== LOAD MODEL ARCHITECTURE
#==============================
try:
self.model= load_model(modelfile)
except Exception as e:
logger.warn("Failed to load model from file %s (err=%s)!" % (modelfile, str(e)))
return -1
if not self.model or self.model is None:
logger.error("Model object is None, loading failed!")
return -1
#===========================
#== SET LOSS & METRICS
#===========================
self.model.compile(optimizer=self.optimizer, loss=self.loss_type, metrics=['accuracy', f1score_metric, precision_metric, recall_metric], run_eagerly=True)
# - Print and draw model
self.model.summary()
plot_model(self.model,to_file='model.png',show_shapes=True)
return 0
def run(self, img_group_1, img_group_2):
""" Compute spectral index """
# - Read image data
if self.__read_imgs() < 0:
logger.error("Failed to read input imgs!")
return -1
# - Check data integrity
good_data = self.__has_good_data(check_mask=True,
check_bad=True,
check_neg=True,
check_same=True)
if not good_data:
logger.warn("Source data selected as bad, skip this source...")
return -1
# - Compute spectral index
if self.__compute_spectral_index(img_group_1, img_group_2) < 0:
logger.error("Failed to compute spectral index (see logs)!")
return -1
# - Fill dict data
self.__fill_data()
return 0
def subtract_bkg(self, bkgs, subtract_only_refch=False):
""" Subtract image background """
if len(bkgs)!=self.nchannels:
logger.error("Number of input bkgs != nchannels, cannot subtract bkg!")
return -1
# - Subtract bkg
if subtract_only_refch:
data= self.img_data[self.refch]
mask= self.img_data_mask[self.refch]
bkg= bkgs[self.refch]
self.img_data[self.refch]-= bkg
self.img_data[self.refch][mask==0]= 0
else:
for i in range(self.nchannels):
data= self.img_data[i]
mask= self.img_data_mask[i]
bkg= bkgs[i]
self.img_data[i]-= bkg
self.img_data[i][mask==0]= 0
# - Draw data & masks?
if self.draw:
fig, axs = plt.subplots(2, self.nchannels)
for i in range(self.nchannels):
axs[0, i].imshow(self.img_data[i])
axs[1, i].imshow(self.img_data_mask[i])
plt.show()
return 0
def read_regions(self):
""" Read regions """
# - Read regions
logger.info("[PROC %d] Reading DS9 region file %s ..." % (procId, self.regionfile))
ret= Utils.read_regions([self.regionfile])
if ret is None:
logger.error("[PROC %d] Failed to read regions (check format)!" % (procId))
return -1
regs= ret[0]
snames= ret[1]
slabels= ret[2]
# - Select region by tag
regs_sel= regs
snames_sel= snames
slabels_sel= slabels
if self.filter_regions_by_tags and self.tags:
logger.info("[PROC %d] Selecting DS9 region with desired tags ..." % (procId))
regs_sel, snames_sel, slabels_sel= Utils.select_regions(regs, self.tags)
if not regs_sel:
logger.warn("[PROC %d] No region left for processing (check input region file)!" % (procId))
return -1
self.sname_label_map= {}
for i in range(len(snames_sel)):
sname= snames_sel[i]
slabel= slabels_sel[i]
self.sname_label_map[sname]= slabel
print("sname_label_map")
print(self.sname_label_map)
# - Compute centroids & radius
centroids, radii= Utils.compute_region_info(regs_sel)
# - Assign sources to each processor
self.nsources= len(regs_sel)
source_indices= list(range(0, self.nsources))
source_indices_split= np.array_split(source_indices, nproc)
source_indices_proc= list(source_indices_split[procId])
self.nsources_proc= len(source_indices_proc)
imin= source_indices_proc[0]
imax= source_indices_proc[self.nsources_proc-1]
self.snames_proc= snames_sel[imin:imax+1]
self.slabels_proc= slabels_sel[imin:imax+1]
self.regions_proc= regs_sel[imin:imax+1]
self.centroids_proc= centroids[imin:imax+1]
self.radii_proc= radii[imin:imax+1]
logger.info("[PROC %d] #%d sources assigned to this processor ..." % (procId, self.nsources_proc))
print("snames_proc %d" % (procId))
print(self.snames_proc)
return 0
def run_ae_reconstruction(self, datalist):
""" Run AE reconstruction """
if procId==MASTER:
aereco_status= 0
# - Create data loader
dl= DataLoader(filename=datalist)
# - Read datalist
logger.info("[PROC %d] Reading datalist %s ..." % (procId, datalist))
dataread_status= dl.read_datalist()
if dataread_status<0:
logger.error("[PROC %d] Failed to read input datalist %s" % (procId, datalist))
aereco_status= -1
else:
# - Run AE reco
logger.info("[PROC %d] Running autoencoder classifier reconstruction ..." % (procId))
ae= FeatExtractorAE(dl)
ae.resize= self.resize_img
ae.set_image_size(self.nx, self.ny)
ae.normalize= self.normalize_img
ae.scale_to_abs_max= self.scale_img_to_abs_max
ae.scale_to_max= self.scale_img_to_max
ae.log_transform_img= self.log_transform_img
ae.scale_img= self.scale_img
ae.scale_img_factors= self.scale_img_factors
ae.standardize_img= self.standardize_img
ae.img_means= self.img_means
ae.img_sigmas= self.img_sigmas
ae.chan_divide= self.img_chan_divide
ae.chan_mins= self.img_chan_mins
ae.erode= self.img_erode
ae.erode_kernel= self.img_erode_kernel
ae.add_channorm_layer= self.add_channorm_layer
aereco_status= ae.reconstruct_data(
self.modelfile_encoder, self.weightfile_encoder,
self.modelfile_decoder, self.weightfile_decoder,
winsize= self.winsize,
outfile_metrics=self.outfile_aerecometrics,
save_imgs= False
)
else:
aereco_status= 0
if comm is not None:
aereco_status= comm.bcast(aereco_status, root=MASTER)
if aereco_status<0:
logger.error("[PROC %d] Failed to run autoencoder reconstruction on data %s, exit!" % (procId, datalist))
return -1
return 0
def __merge_data(self, dlist):
""" Merge feature data """
# - Check input list
if not dlist:
logger.error("Empty data dict list given!")
return -1
for i in range(len(dlist)):
d = dlist[i]
if not d:
logger.error("Data dict %d is empty!" % (i + 1))
return -1
# - Compute number of vars
nvars_tot = 0
for d in dlist:
print("d")
print(d)
nentries = len(d.keys())
firstitem = next(iter(d.items()))
nvars = len(firstitem[1].keys()) - 2
nvars_tot += nvars
logger.info("Data dict has #%d entries (#%d vars) ..." %
(nentries, nvars))
logger.info("Merged set is expected to have %d vars ..." % (nvars_tot))
# - Merge features
logger.info("Merging feature data for input data dict ...")
dmerged = collections.OrderedDict()
for d in dlist:
for key, value in d.items():
if key not in dmerged:
dmerged[key] = collections.OrderedDict({})
dmerged[key].update(value)
dmerged[key].move_to_end("id")
# - Remove rows with less number of entries
logger.info("Removing rows with number of vars !=%d ..." % (nvars_tot))
self.par_dict_list = []
for key, value in dmerged.items():
nvars = len(value.keys()) - 2
if nvars != nvars_tot:
logger.info(
"Removing entry (%s) as number of vars (%d) is !=%d ..." %
(key, nvars, nvars_tot))
#del dmerged[key]
continue
self.par_dict_list.append(value)
return 0
def __predict(self):
#====================================================
#== CHECK DATA & MODEL
#====================================================
# - Check if data are set
if self.data is None:
logger.error("Input data array is None!")
return -1
# - Check if clustering model is set
if self.clusterer is None:
logger.error("Clusterer is not set!")
return -1
# - Retrieve prediction data from current model
logger.info(
"Retrieving prediction data from current model (if any) ...")
self.prediction_data = self.clusterer.prediction_data_
#====================================================
#== CLUSTER DATA USING SAVED MODEL
#====================================================
logger.info("Encode input data using loaded model ...")
self.labels, self.probs = hdbscan.approximate_predict(
self.clusterer, self.data)
#================================
#== SAVE CLUSTERED DATA
#================================
logger.info("Saving unsupervised encoded data to file ...")
N = self.data.shape[0]
print("Cluster data N=", N)
snames = np.array(self.source_names).reshape(N, 1)
objids = np.array(self.data_classids).reshape(N, 1)
clustered_data = np.concatenate(
(snames, objids, self.labels, self.probs), axis=1)
head = "# sname id clustid clustprob"
Utils.write_ascii(clustered_data, self.outfile, head)
#================================
#== PLOT
#================================
logger.info("Plotting results ...")
self.__plot_predict(self.clusterer, self.data, self.labels,
self.source_names, self.data_labels,
self.prediction_data, self.prediction_extra_data,
self.outfile_plot)
return 0
def set_data_from_file(self, filename):
""" Set data from input file. Expected format: sname, N features, classid """
# - Read table
row_start = 0
try:
table = ascii.read(filename, data_start=row_start)
except:
logger.error("Failed to read feature file %s!" % filename)
return -1
print(table.colnames)
print(table)
ncols = len(table.colnames)
nfeat = ncols - 2
# - Set data vectors
rowIndex = 0
self.data_labels = []
self.data_classids = []
self.source_names = []
featdata = []
for data in table:
sname = data[0]
classid = data[ncols - 1]
label = self.classid_label_map[classid]
self.source_names.append(sname)
self.data_labels.append(label)
self.data_classids.append(classid)
featdata_curr = []
for k in range(nfeat):
featdata_curr.append(data[k + 1])
featdata.append(featdata_curr)
self.data = np.array(featdata)
if self.data.size == 0:
logger.error("Empty feature data vector read!")
return -1
self.nsamples = data_shape[0]
self.nfeatures = data_shape[1]
logger.info("#nsamples=%d" % (self.nsamples))
# - Set pre-classified data
logger.info("Setting pre-classified data (if any) ...")
self.__set_preclass_data()
return 0
def __read_sdata(self, index):
""" Read source data """
# - Check index
if index<0 or index>=self.datasize:
logger.error("Invalid index %d given!" % index)
return None
# - Init sdata
sdata= SData()
sdata.refch= self.refch
sdata.kernsize= self.kernsize
sdata.draw= self.draw
sdata.save_ssim_pars= self.save_ssim_pars
sdata.negative_pix_fract_thr= self.negative_pix_fract_thr
sdata.bad_pix_fract_thr= self.bad_pix_fract_thr
sdata_mask= SData()
sdata_mask.refch= self.refch
sdata_mask.kernsize= self.kernsize
sdata_mask.draw= self.draw
sdata_mask.save_ssim_pars= self.save_ssim_pars
sdata_mask.negative_pix_fract_thr= self.negative_pix_fract_thr
sdata_mask.bad_pix_fract_thr= self.bad_pix_fract_thr
# - Read source image data
logger.debug("Reading source image data %d ..." % index)
#d= self.datalist["data"][index]
d= self.datalist[index]
if sdata.set_from_dict(d)<0:
logger.error("Failed to set source image data %d!" % index)
return None
if sdata.read_imgs()<0:
logger.error("Failed to read source images %d!" % index)
return None
# - Read source masked image data
logger.debug("Reading source masked image data %d ..." % index)
#d= self.datalist_mask["data"][index]
d= self.datalist_mask[index]
if sdata_mask.set_from_dict(d)<0:
logger.error("Failed to set source masked image data %d!" % index)
return None
if sdata_mask.read_imgs()<0:
logger.error("Failed to read source masked images %d!" % index)
return None
return sdata, sdata_mask
def __create_pipeline(self):
""" Build the feature selector pipeline """
# - Create classifier inventory
logger.info("Creating classifier inventory ...")
self.__create_classifier_inventory()
# - Set min/max nfeat range
nf_min = self.nfeat_min
nf_max = self.nfeat_max
if nf_max == -1:
nf_max = self.nfeatures
self.nfeats = []
for i in range(nf_min, nf_max + 1):
self.nfeats.append(i)
# - Create models
self.model = self.__create_model()
if self.model is None:
logger.error("Created model is None!")
return -1
for i in range(len(self.nfeats)):
m = self.__create_model()
self.models.append(m)
# - Define dataset split (unique for all models)
self.cv = StratifiedKFold(n_splits=self.cv_nsplits,
shuffle=True,
random_state=self.cv_seed)
# - Create RFE & pipeline
self.rfe = RFECV(
estimator=self.model,
step=1,
#cv=self.cv,
min_features_to_select=self.nfeat_min)
self.pipeline = Pipeline(steps=[('featsel',
self.rfe), ('model', self.model)])
for i in range(len(self.nfeats)):
n = self.nfeats[i]
r = RFE(
estimator=self.models[i],
#cv=self.cv,
n_features_to_select=n)
p = Pipeline(steps=[('featsel', r), ('model', self.models[i])])
self.pipelines.append(p)
return 0
def run_from_dictlist(self, dlist, outfile='featdata_merged.dat'):
""" Run feature merger """
# - Read feature data and merge
logger.info("Merging input feature data dicts ...")
if self.__merge_data(dlist) < 0:
logger.error("Failed to merge data!")
return -1
# - Save data
logger.info("Saving merged data to file %s ..." % (outfile))
self.__save(outfile)
return 0
def run_predict(self, datafile, modelfile):
""" Run predict using input dataset """
#================================
#== LOAD DATA
#================================
# - Check inputs
if datafile == "":
logger.error("Empty data file specified!")
return -1
if self.set_data_from_file(datafile) < 0:
logger.error("Failed to read datafile %s!" % datafile)
return -1
#================================
#== LOAD MODEL
#================================
logger.info("Loading the UMAP reducer from file %s ..." % modelfile)
try:
self.reducer = pickle.load((open(modelfile, 'rb')))
except Exception as e:
logger.error("Failed to load model from file %s!" % (modelfile))
return -1
#================================
#== PREDICT
#================================
if self.__predict() < 0:
logger.error("Predict failed!")
return -1
return 0
def classify_sources(self):
""" Run source classification """
# - Run source classification
if procId==MASTER:
sclass_status= 0
# - Define sclassifier class
multiclass= True
if self.binary_class:
multiclass= False
sclass= SClassifier(multiclass=multiclass)
sclass.normalize= self.normalize_feat
sclass.outfile= self.outfile_sclass
sclass.outfile_metrics= self.outfile_sclass_metrics
sclass.outfile_cm= self.outfile_sclass_cm
sclass.outfile_cm_norm= self.outfile_sclass_cm_norm
sclass.save_labels= self.save_class_labels
sclass.find_outliers= self.find_outliers
sclass.outlier_modelfile= self.modelfile_outlier
sclass.outlier_thr= self.anomaly_thr
sclass.outlier_max_samples= self.max_samples
sclass.outlier_max_features= self.max_features
sclass.save_outlier= self.save_outlier
sclass.outlier_outfile= self.outfile_outlier
# - Run classification
sclass_status= sclass.run_predict(
#data=self.feat_colors, class_ids=self.feat_colors_classids, snames=self.feat_colors_snames,
data=self.feat_all, class_ids=self.feat_all_classids, snames=self.feat_all_snames,
modelfile=self.modelfile,
scalerfile=self.scalerfile
)
if sclass_status<0:
logger.error("[PROC %d] Failed to run classifier on data %s!" % (procId, featfile_allfeat))
else:
sclass_status= 0
if comm is not None:
sclass_status= comm.bcast(sclass_status, root=MASTER)
if sclass_status<0:
logger.error("[PROC %d] Failed to run classifier on data %s, exit!" % (procId, featfile_allfeat))
return -1
return 0
def run(self, datafile):
""" Run feature selection """
#================================
#== LOAD DATA
#================================
# - Check inputs
if datafile == "":
logger.error("Empty data file specified!")
return -1
if self.set_data_from_file(datafile) < 0:
logger.error("Failed to read datafile %s!" % datafile)
return -1
#================================
#== EVALUATE MODELS
#================================
logger.info("Evaluating models ...")
if self.__evaluate_model() < 0:
logger.error("Failed to evaluate models!")
return -1
#================================
#== SAVE
#================================
logger.info("Saving results ...")
if self.__save() < 0:
logger.error("Failed to save results!")
return -1
return 0
def run(self, data, class_ids=[], snames=[]):
""" Run feature selection using input dataset """
#================================
#== LOAD DATA
#================================
# - Check inputs
if data is None:
logger.error("None input data specified!")
return -1
if self.set_data(data, class_ids, snames) < 0:
logger.error("Failed to read datafile %s!" % datafile)
return -1
#================================
#== EVALUATE MODELS
#================================
logger.info("Evaluating models ...")
if self.__evaluate_model() < 0:
logger.error("Failed to evaluate models!")
return -1
#================================
#== SAVE
#================================
logger.info("Saving results ...")
if self.__save() < 0:
logger.error("Failed to save results!")
return -1
return 0
def run_predict(self, data, class_ids=[], snames=[], modelfile=''):
""" Run precit using input dataset """
#================================
#== LOAD DATA
#================================
# - Check inputs
if data is None:
logger.error("None input data specified!")
return -1
if self.set_data(data, class_ids, snames) < 0:
logger.error("Failed to read datafile %s!" % datafile)
return -1
#================================
#== LOAD MODEL
#================================
logger.info("Loading the clustering model from file %s ..." %
modelfile)
try:
self.clusterer, self.prediction_extra_data = pickle.load(
(open(modelfile, 'rb')))
except Exception as e:
logger.error("Failed to load model from file %s!" % (modelfile))
return -1
#================================
#== PREDICT
#================================
if self.__predict() < 0:
logger.error("Predict failed!")
return -1
return 0
def select(self, data, selcols, class_ids=[], snames=[]):
""" Select data columns provided in selcols list """
#================================
#== LOAD DATA
#================================
# - Check inputs
if data is None:
logger.error("None input data specified!")
return -1
if self.set_data(data, class_ids, snames) < 0:
logger.error("Failed to read datafile %s!" % datafile)
return -1
#================================
#== SELECT COLUMNS
#================================
logger.info("Extracting columns ...")
if self.__select_cols(selcols) < 0:
logger.error("Failed to select data columns!")
return -1
#================================
#== SAVE
#================================
logger.info("Saving results ...")
if self.__save() < 0:
logger.error("Failed to save results!")
return -1
return 0
def select_from_file(self, datafile, selcols):
""" Select data columns provided in selcols list """
#================================
#== LOAD DATA
#================================
# - Check inputs
if datafile == "":
logger.error("Empty data file specified!")
return -1
if self.set_data_from_file(datafile) < 0:
logger.error("Failed to read datafile %s!" % datafile)
return -1
#================================
#== SELECT COLUMNS
#================================
logger.info("Extracting columns ...")
if self.__select_cols(selcols) < 0:
logger.error("Failed to select data columns!")
return -1
#================================
#== SAVE
#================================
logger.info("Saving results ...")
if self.__save() < 0:
logger.error("Failed to save results!")
return -1
return 0
def shrink_masks(self, kernsizes=[]):
""" Shrink masks """
# - Set erosion kernel sizes
if not kernsizes or len(kernsizes)!=self.nchannels:
logger.info("kernsizes not specified, setting kernsize=%d for all channels ..." % (self.kernsize))
kernsizes= [self.kernsize]*self.nchannels
print("--> kernsizes")
print(kernsizes)
# - Erode masks
if self.draw:
fig, axs = plt.subplots(4, self.nchannels)
try:
counter= 0
for i in range(self.nchannels):
data= np.copy(self.img_data[i])
mask= np.copy(self.img_data_mask[i])
# - Do erosion if kernsize is >0
if kernsizes[i]>0:
structel= cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (kernsizes[i],kernsizes[i]))
mask_eroded = cv2.erode(mask, structel, iterations = 1)
self.img_data_mask[i]= mask_eroded
data_eroded= self.img_data[i]
data_eroded[mask_eroded==0]= 0
self.img_data[i]= data_eroded
if self.draw:
axs[0, i].imshow(data)
counter+= 1
axs[1, i].imshow(mask)
counter+= 1
axs[2, i].imshow(self.img_data[i])
counter+= 1
axs[3, i].imshow(self.img_data_mask[i])
counter+= 1
except Exception as e:
logger.error("Failed to shrink masks (err=%s)!" % (str(e)))
return -1
if self.draw:
plt.show()
return 0
def __predict(self):
#====================================================
#== CHECK DATA & MODEL
#====================================================
# - Check if data are set
if self.data is None:
logger.error("Input data array is None!")
return -1
# - Check if reducer is set
if self.reducer is None:
logger.error("UMAP reducer is not set!")
return -1
#====================================================
#== ENCODE DATA
#====================================================
logger.info("Encode input data using loaded model ...")
self.encoded_data_unsupervised = self.reducer.transform(self.data)
#================================
#== SAVE ENCODED DATA
#================================
# - Unsupervised encoded data
logger.info("Saving unsupervised encoded data to file ...")
N = self.encoded_data_unsupervised.shape[0]
print("Unsupervised encoded data shape=",
self.encoded_data_unsupervised.shape)
print("Unsupervised encoded data N=", N)
snames = np.array(self.source_names).reshape(N, 1)
objids = np.array(self.data_ids).reshape(N, 1)
# - Save unsupervised encoded data
enc_data = np.concatenate(
(snames, self.encoded_data_unsupervised, objids), axis=1)
znames_counter = list(range(1, self.encoded_data_dim + 1))
znames = '{}{}'.format('z',
' z'.join(str(item) for item in znames_counter))
head = '{} {} {}'.format("# sname", znames, "id")
Utils.write_ascii(enc_data, self.outfile_encoded_data_unsupervised,
head)
return 0