def test_layers_activation_for_voxel_picture(self):
pic = get_test_voxel_picture()
ds = pscn.SparseDataset('test_ds', 'UNLABELEDBATCH', 1, 40)
ds.add_voxel_picture(pic)
net = create_dC2()
lois = net.layer_activations_for_dataset(ds)
self.assertEqual(len(lois), 19)
def test_batch_processing_for_voxel_picture(self):
pic = get_test_voxel_picture()
ds = pscn.SparseDataset('test_ds', 'TRAINBATCH', 1, 40)
ds.add_voxel_picture(pic)
net = create_dC2()
batch_gen = net.batch_generator(ds, 1)
batch = next(batch_gen)
batch_output = net.processBatchForward(batch)
self.assertEqual(batch_output['spatialSize'], 1)
def learn_simple_network(full=False, batchSize=10, limit=1, epoch=2):
network = create_dC2()
print("Created network")
dataset = generate_modelnet_dataset(full=full, limit=limit)
dataset.summary()
print("Created dataset {0}".format(dataset.name))
for epoch in xrange(1, epoch):
learning_rate = 0.003 * np.exp(-0.05 / 2 * epoch)
# print("epoch {0}, lr={1} ".format(epoch, learning_rate), end='')
network.processDataset(dataset, batchSize=batchSize,
learningRate=learning_rate)
def test_predict(self):
unlabeled_dataset = SparseDataset("One pic", 'UNLABELEDBATCH', 1, 1)
network = create_dC2()
num_of_inputs = 5
nClasses = 40
renderSize = 40
test_file = ('SparseConvNet/Data/ModelNet/night_stand/'
'train/night_stand_0180.off')
for i in range(num_of_inputs):
unlabeled_dataset.add_picture(Off3DPicture(test_file, renderSize))
matrix_of_preds = network.predict(unlabeled_dataset)
self.assertEqual(matrix_of_preds.shape, (num_of_inputs, nClasses))
def learn_simple_network(full=False, batchSize=10, limit=1, epoch=2):
network = create_dC2()
print("Created network")
dataset = generate_modelnet_dataset(full=full, limit=limit)
dataset.summary()
print("Created dataset {0}".format(dataset.name))
for epoch in xrange(1, epoch):
learning_rate = 0.003 * np.exp(-0.05 / 2 * epoch)
# print("epoch {0}, lr={1} ".format(epoch, learning_rate), end='')
network.processDataset(dataset,
batchSize=batchSize,
learningRate=learning_rate)
def test_predict(self):
unlabeled_dataset = SparseDataset("One pic", 'UNLABELEDBATCH', 1, 1)
network = create_dC2()
num_of_inputs = 5
nClasses = 40
renderSize = 40
test_file = ('SparseConvNet/Data/ModelNet/night_stand/'
'train/night_stand_0180.off')
for i in range(num_of_inputs):
unlabeled_dataset.add_picture(Off3DPicture(test_file, renderSize))
matrix_of_preds = network.predict(unlabeled_dataset)
self.assertEqual(matrix_of_preds.shape, (num_of_inputs, nClasses))
def test_layer_activation(self):
network = create_dC2()
network.loadWeights('SparseConvNet/weights/ModelNet_10_repeat_bs100_nthrd10/ModelNet', 200)
lois = [
network.layer_activations(
Off3DPicture(
'SparseConvNet/Data/ModelNet/car/test/car_0216.off', 40)),
network.layer_activations(
Off3DPicture(
'SparseConvNet/Data/ModelNet/sink/test/sink_0133.off', 40))
]
self.assertEqual(len(lois[0]), 19)
def test_layer_activation(self):
network = create_dC2()
network.loadWeights(
'SparseConvNet/weights/ModelNet_10_repeat_bs100_nthrd10/ModelNet',
200)
lois = [
network.layer_activations(
Off3DPicture(
'SparseConvNet/Data/ModelNet/car/test/car_0216.off', 40)),
network.layer_activations(
Off3DPicture(
'SparseConvNet/Data/ModelNet/sink/test/sink_0133.off', 40))
]
self.assertEqual(len(lois[0]), 19)
def test_dC2_creation_and_loading(self):
network = create_dC2()
self.assertEqual(type(network), SparseNetwork)
layers = load_and_get_weights(network)
self.assertEqual(len(layers), 18)
def test_dC2_creation_and_loading(self):
network = create_dC2()
self.assertEqual(type(network), SparseNetwork)
layers = load_and_get_weights(network)
self.assertEqual(len(layers), 18)