ssvm = learners.NSlackSSVM(
model, verbose=2, C=1, max_iter=1000, n_jobs=-1,
tol=0.0001, show_loss_every=5,
inactive_threshold=1e-3, inactive_window=10, batch_size=100)
ssvm.fit(X_valid, valid_Y)
print ssvm.score(X_valid,valid_Y)
print ssvm.score(X_test,test_Y)
predict = ssvm.predict(X_valid)
for i in range(0, len(X_valid)):
predict_result = predict[i]
fe = Feature()
x = glob.glob("../data_road/training/image/um_000001.png")
print len(x)
fe.loadImage(x[0])
fe.loadSuperpixelImage()
image = fe.getImage()
superpixels = valid_superpixels[i][0]
newIm = np.zeros((image.shape[0], image.shape[1], image.shape[2]))
numSuperpixels = np.max(superpixels)+1
for i in xrange(0,numSuperpixels):
indices = np.where(superpixels==i)
prediction = predict_result[i]
image[indices] = 1
sp.showPlots("im_name", image, numSuperpixels, superpixels)
#superpixels =
#sp.showPlots(x, y_pred[0], np.max(superpixels),superpixels):
#print y_pred
# we throw away void superpixels and flatten everything
valid_pixels_labels = []
test_pixels_labels = []
valid_files = []
test_files = []
valid_files_count = 0
test_files_count = 0
valid_superpixels = []
validationOriginalImage = []
test_superpixels = []
testOriginalImage = []
train_superpixels = []
for i in xrange(0,num_files):
fe = Feature()
fe.loadImage(im_file_names[i])
fe.loadSuperpixelImage()
#fe.loadSuperpixelFromFile(sp_file_names[i])
fe.loadLabelImage(label_file_names[i])
featureVectors = fe.getFeaturesVectors()
labels = fe.getSuperPixelLabels()
#Test purposes
edges, edgeFeatures1, edgeFeatures2 = fe.getEdges()
if file_labels[i] != TESTING_LABEL:
# store data
if file_labels[i] == TRAINING_LABEL:
train_edges.append(edges)
train_edgesFeatures1.append(edgeFeatures1)
valid_edgesFeatures2 = []
valid_pixels_labels = []
test_pixels_labels = []
valid_files = []
test_files = []
valid_files_count = 0
test_files_count = 0
valid_superpixels = []
validationOriginalImage = []
test_superpixels = []
testOriginalImage = []
train_superpixels = []
for i in xrange(0, num_files):
fe = Feature()
fe.loadImage(im_file_names[i])
fe.loadSuperpixelImage()
#fe.loadSuperpixelFromFile(sp_file_names[i])
fe.loadLabelImage(label_file_names[i])
featureVectors = fe.getFeaturesVectors()
labels = fe.getSuperPixelLabels()
#Test purposes
edges, edgeFeatures1, edgeFeatures2 = fe.getEdges()
if file_labels[i] != TESTING_LABEL:
# store data
if file_labels[i] == TRAINING_LABEL:
train_edges.append(edges)
train_edgesFeatures1.append(edgeFeatures1)
tol=0.0001,
show_loss_every=5,
inactive_threshold=1e-3,
inactive_window=10,
batch_size=100)
ssvm.fit(X_valid, valid_Y)
print ssvm.score(X_valid, valid_Y)
print ssvm.score(X_test, test_Y)
predict = ssvm.predict(X_valid)
for i in range(0, len(X_valid)):
predict_result = predict[i]
fe = Feature()
x = glob.glob("../data_road/training/image/um_000001.png")
print len(x)
fe.loadImage(x[0])
fe.loadSuperpixelImage()
image = fe.getImage()
superpixels = valid_superpixels[i][0]
newIm = np.zeros((image.shape[0], image.shape[1], image.shape[2]))
numSuperpixels = np.max(superpixels) + 1
for i in xrange(0, numSuperpixels):
indices = np.where(superpixels == i)
prediction = predict_result[i]
image[indices] = 1
sp.showPlots("im_name", image, numSuperpixels, superpixels)
#superpixels =
#sp.showPlots(x, y_pred[0], np.max(superpixels),superpixels):
#print y_pred
# we throw away void superpixels and flatten everything