df = stack.to_pandas()
# df = stack.to_pandas(res=500)
df = df.melt(id_vars=['x', 'y'])
from plotnine import *
(ggplot(df, aes(x="x", y="y", fill="value")) + geom_tile() + coord_fixed() +
facet_wrap("variable") + theme_light() + theme(axis_title=element_blank()))
from sklearn.ensemble import RandomForestClassifier
clf = RandomForestClassifier(n_estimators=100)
clf.fit(X, y)
stack.predict(clf, output='test', overwrite=True, height=25)
stack.predict_proba(clf, output='test', overwrite=True, height=25)
test = RasterRow('test')
from grass.pygrass.modules.shortcuts import raster as r
r.colors('test', color='random')
test
test.close()
from sklearn.model_selection import cross_validate
cross_validate(clf, X, y, cv=3)
from grass.pygrass.gis.region import Region
from grass.pygrass.modules.grid.grid import GridModule
from grass.pygrass.modules.grid import split
from grass.pygrass.modules.shortcuts import general as g
def main():
try:
import sklearn
import joblib
if sklearn.__version__ < "0.20":
gs.fatal(
"Package python3-scikit-learn 0.20 or newer is not installed")
except ImportError:
gs.fatal("Package python3-scikit-learn 0.20 or newer is not installed")
# parser options
group = options["group"]
output = options["output"]
model_load = options["load_model"]
probability = flags["p"]
prob_only = flags["z"]
chunksize = int(options["chunksize"])
# remove @ from output in case overwriting result
if "@" in output:
output = output.split("@")[0]
# check probabilities=True if prob_only=True
if prob_only is True and probability is False:
gs.fatal("Need to set probabilities=True if prob_only=True")
# reload fitted model and training data
estimator, y, class_labels = joblib.load(model_load)
# define RasterStack
stack = RasterStack(group=group)
# perform raster prediction
region = Region()
row_incr = math.ceil(chunksize / region.cols)
# do not read by increments if increment > n_rows
if row_incr >= region.rows:
row_incr = None
# prediction
if prob_only is False:
gs.message("Predicting classification/regression raster...")
stack.predict(
estimator=estimator,
output=output,
height=row_incr,
overwrite=gs.overwrite(),
)
if probability is True:
gs.message("Predicting class probabilities...")
stack.predict_proba(
estimator=estimator,
output=output,
class_labels=np.unique(y),
overwrite=gs.overwrite(),
height=row_incr,
)
# assign categories for classification map
if class_labels is not None and prob_only is False:
rules = []
for val, lab in class_labels.items():
rules.append(",".join([str(val), lab]))
rules = "\n".join(rules)
rules_file = string_to_rules(rules)
r.category(map=output, rules=rules_file, separator="comma")
def main():
try:
import sklearn
from sklearn.externals import joblib
if sklearn.__version__ < '0.20':
gs.fatal("Scikit learn 0.20 or newer is required")
except ImportError:
gs.fatal("Scikit learn 0.20 or newer is not installed")
# -------------------------------------------------------------------------
# Parser options
# -------------------------------------------------------------------------
group = options['group']
output = options['output']
model_load = options['load_model']
probability = flags['p']
prob_only = flags['z']
chunksize = int(options['chunksize'])
# remove @ from output in case overwriting result
if '@' in output:
output = output.split('@')[0]
# check that probabilities=True if prob_only=True
if prob_only is True and probability is False:
gs.fatal('Need to set probabilities=True if prob_only=True')
# -------------------------------------------------------------------------
# Reload fitted model and trainign data
# -------------------------------------------------------------------------
X, y, sample_coords, group_id, estimator = joblib.load(model_load)
# -------------------------------------------------------------------------
# Define RasterStack
# -------------------------------------------------------------------------
# fetch individual raster names from group
maplist = gs.read_command("i.group", group=group,
flags="g").split(os.linesep)[:-1]
# create RasterStack
stack = RasterStack(rasters=maplist)
# -------------------------------------------------------------------------
# Perform raster prediction
# -------------------------------------------------------------------------
# calculate chunksize
row = stack.read(1)
rowsize_mg = row.nbytes * 1e-6
row_incr = int(float(chunksize) / float(rowsize_mg))
# prediction
if prob_only is False:
gs.message('Predicting classification/regression raster...')
stack.predict(estimator=estimator,
output=output,
height=row_incr,
overwrite=gs.overwrite())
if probability is True:
gs.message('Predicting class probabilities...')
stack.predict_proba(estimator=estimator,
output=output,
class_labels=np.unique(y),
overwrite=gs.overwrite(),
height=row_incr)