def main():
model = KMeansThresholding()
train_files_x = find_files("data/train")
train_files_y = find_files("data/train_cleaned")
y_pred = map(lambda x: extract_x([x],0), train_files_x)
y_pred = list(map(predict, y_pred))
y_pred = np.concatenate(y_pred)
y_true = np.concatenate(list(map(lambda x: extract_y([x],0) ,train_files_y)))
print("Calculating MSE")
mse = metrics.mean_squared_error(y_true, y_pred)
print("MSE: %0.2f (+/- %0.2f)" % (mse.mean()*100, mse.std() * 200))
def main():
r = int(getarg(1, 0))
k = int(getarg(2, 0))
print("Extracting with radius=%d" % (r))
prefix = "data/train"
if(k > 0):
prefix += "_k"+str(k)
train_files_x = find_files(prefix)
X = extract_x(train_files_x,r)
train_files_y = find_files("data/train_cleaned")
y = extract_y(train_files_y,r)
outname = prefix+"_r"+str(r)+".npz"
np.savez(outname,X=X,y=y)
def main():
r = int(getarg(1, 0))
k = int(getarg(2, 3))
print("Extracting with radius=%d" % (r))
prefix = "data/train"
outname = prefix+"_c_k"+str(k)+"_r"+str(r)+".npz"
print("Saving to: "+ outname)
train_files_x = find_files(prefix)
X = extract_x(train_files_x,r)
Xk = extract_x(find_files(prefix+"_k"+str(k)),r,True)
X = np.concatenate([X,Xk], axis=1)
train_files_y = find_files("data/train_cleaned")
y = extract_y(train_files_y,r)
np.savez(outname,X=X,y=y)
def main():
test_files = find_files("data/test")
model = pickle.load(open("data/raw_lin.p", "rb"))
for f in test_files:
w,h = Image.open(f).size
X = extract_x([f])
y = np.reshape(model.predict(X), (h,w))
y = np.vectorize(filter_point_value)(y)
img = Image.fromarray(y.astype(np.uint8))
outf = str.replace(f,"data/test","out/raw_lin")
img.save(outf)