def test_create_ssd_mobilenet_v2_model_from_config(self):
model_text_proto = """
ssd {
feature_extractor {
type: 'ssd_mobilenet_v2'
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
box_coder {
faster_rcnn_box_coder {
}
}
matcher {
argmax_matcher {
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
ssd_anchor_generator {
aspect_ratios: 1.0
}
}
image_resizer {
fixed_shape_resizer {
height: 320
width: 320
}
}
box_predictor {
convolutional_box_predictor {
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
}
normalize_loc_loss_by_codesize: true
loss {
classification_loss {
weighted_softmax {
}
}
localization_loss {
weighted_smooth_l1 {
}
}
}
weight_regression_loss_by_score: true
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
model = self.create_model(model_proto)
self.assertIsInstance(model, ssd_meta_arch.SSDMetaArch)
self.assertIsInstance(model._feature_extractor,
SSDMobileNetV2FeatureExtractor)
self.assertTrue(model._normalize_loc_loss_by_codesize)
self.assertTrue(model._target_assigner._weight_regression_loss_by_score)
def test_create_ssd_inception_v2_model_from_config(self):
model_text_proto = """
ssd {
feature_extractor {
type: 'ssd_inception_v2'
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
override_base_feature_extractor_hyperparams: true
}
box_coder {
faster_rcnn_box_coder {
}
}
matcher {
argmax_matcher {
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
ssd_anchor_generator {
aspect_ratios: 1.0
}
}
image_resizer {
fixed_shape_resizer {
height: 320
width: 320
}
}
box_predictor {
convolutional_box_predictor {
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
}
loss {
classification_loss {
weighted_softmax {
}
}
localization_loss {
weighted_smooth_l1 {
}
}
}
use_expected_classification_loss_under_sampling: true
minimum_negative_sampling: 10
desired_negative_sampling_ratio: 2
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
model = self.create_model(model_proto)
self.assertIsInstance(model, ssd_meta_arch.SSDMetaArch)
self.assertIsInstance(model._feature_extractor,
SSDInceptionV2FeatureExtractor)
self.assertIsNotNone(model._expected_classification_loss_under_sampling)
self.assertEqual(
model._expected_classification_loss_under_sampling.keywords, {
'minimum_negative_sampling': 10,
'desired_negative_sampling_ratio': 2
})
def test_create_rfcn_resnet_v1_model_from_config(self):
model_text_proto = """
faster_rcnn {
num_classes: 3
image_resizer {
keep_aspect_ratio_resizer {
min_dimension: 600
max_dimension: 1024
}
}
feature_extractor {
type: 'faster_rcnn_resnet101'
}
first_stage_anchor_generator {
grid_anchor_generator {
scales: [0.25, 0.5, 1.0, 2.0]
aspect_ratios: [0.5, 1.0, 2.0]
height_stride: 16
width_stride: 16
}
}
first_stage_box_predictor_conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
initial_crop_size: 14
maxpool_kernel_size: 2
maxpool_stride: 2
second_stage_box_predictor {
rfcn_box_predictor {
conv_hyperparams {
op: CONV
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
}
second_stage_post_processing {
batch_non_max_suppression {
score_threshold: 0.01
iou_threshold: 0.6
max_detections_per_class: 100
max_total_detections: 300
}
score_converter: SOFTMAX
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
for extractor_type, extractor_class in FRCNN_RESNET_FEAT_MAPS.items():
model_proto.faster_rcnn.feature_extractor.type = extractor_type
model = model_builder.build(model_proto, is_training=True)
self.assertIsInstance(model, rfcn_meta_arch.RFCNMetaArch)
self.assertIsInstance(model._feature_extractor, extractor_class)
def test_create_ssd_resnet_v1_ppn_model_from_config(self):
model_text_proto = """
ssd {
feature_extractor {
type: 'ssd_resnet_v1_50_ppn'
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
box_coder {
mean_stddev_box_coder {
}
}
matcher {
bipartite_matcher {
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
ssd_anchor_generator {
aspect_ratios: 1.0
}
}
image_resizer {
fixed_shape_resizer {
height: 320
width: 320
}
}
box_predictor {
weight_shared_convolutional_box_predictor {
depth: 1024
class_prediction_bias_init: -4.6
conv_hyperparams {
activation: RELU_6,
regularizer {
l2_regularizer {
weight: 0.0004
}
}
initializer {
variance_scaling_initializer {
}
}
}
num_layers_before_predictor: 2
kernel_size: 1
}
}
loss {
classification_loss {
weighted_softmax {
}
}
localization_loss {
weighted_l2 {
}
}
classification_weight: 1.0
localization_weight: 1.0
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
for extractor_type, extractor_class in SSD_RESNET_V1_PPN_FEAT_MAPS.items():
model_proto.ssd.feature_extractor.type = extractor_type
model = model_builder.build(model_proto, is_training=True)
self.assertIsInstance(model, ssd_meta_arch.SSDMetaArch)
self.assertIsInstance(model._feature_extractor, extractor_class)
def testGetContextFeatureLengthFromModelConfig(self):
model_config = model_pb2.DetectionModel()
model_config.faster_rcnn.context_config.context_feature_length = 100
context_feature_length = config_util.get_context_feature_length(
model_config)
self.assertAllEqual(context_feature_length, 100)
def _create_model(
self,
model_fn=ssd_meta_arch.SSDMetaArch,
apply_hard_mining=True,
normalize_loc_loss_by_codesize=False,
add_background_class=True,
random_example_sampling=False,
expected_loss_weights=model_pb2.DetectionModel().ssd.loss.NONE,
min_num_negative_samples=1,
desired_negative_sampling_ratio=3,
use_keras=False,
predict_mask=False,
use_static_shapes=False,
nms_max_size_per_class=5,
calibration_mapping_value=None):
is_training = False
num_classes = 1
mock_anchor_generator = MockAnchorGenerator2x2()
if use_keras:
mock_box_predictor = test_utils.MockKerasBoxPredictor(
is_training, num_classes, add_background_class=add_background_class)
else:
mock_box_predictor = test_utils.MockBoxPredictor(
is_training, num_classes, add_background_class=add_background_class)
mock_box_coder = test_utils.MockBoxCoder()
if use_keras:
fake_feature_extractor = FakeSSDKerasFeatureExtractor()
else:
fake_feature_extractor = FakeSSDFeatureExtractor()
mock_matcher = test_utils.MockMatcher()
region_similarity_calculator = sim_calc.IouSimilarity()
encode_background_as_zeros = False
def image_resizer_fn(image):
return [tf.identity(image), tf.shape(image)]
classification_loss = losses.WeightedSigmoidClassificationLoss()
localization_loss = losses.WeightedSmoothL1LocalizationLoss()
non_max_suppression_fn = functools.partial(
post_processing.batch_multiclass_non_max_suppression,
score_thresh=-20.0,
iou_thresh=1.0,
max_size_per_class=nms_max_size_per_class,
max_total_size=nms_max_size_per_class,
use_static_shapes=use_static_shapes)
score_conversion_fn = tf.identity
calibration_config = calibration_pb2.CalibrationConfig()
if calibration_mapping_value:
calibration_text_proto = """
function_approximation {
x_y_pairs {
x_y_pair {
x: 0.0
y: %f
}
x_y_pair {
x: 1.0
y: %f
}}}""" % (calibration_mapping_value, calibration_mapping_value)
text_format.Merge(calibration_text_proto, calibration_config)
score_conversion_fn = (
post_processing_builder._build_calibrated_score_converter( # pylint: disable=protected-access
tf.identity, calibration_config))
classification_loss_weight = 1.0
localization_loss_weight = 1.0
negative_class_weight = 1.0
normalize_loss_by_num_matches = False
hard_example_miner = None
if apply_hard_mining:
# This hard example miner is expected to be a no-op.
hard_example_miner = losses.HardExampleMiner(
num_hard_examples=None, iou_threshold=1.0)
random_example_sampler = None
if random_example_sampling:
random_example_sampler = sampler.BalancedPositiveNegativeSampler(
positive_fraction=0.5)
target_assigner_instance = target_assigner.TargetAssigner(
region_similarity_calculator,
mock_matcher,
mock_box_coder,
negative_class_weight=negative_class_weight)
model_config = model_pb2.DetectionModel()
if expected_loss_weights == model_config.ssd.loss.NONE:
expected_loss_weights_fn = None
else:
raise ValueError('Not a valid value for expected_loss_weights.')
code_size = 4
kwargs = {}
if predict_mask:
kwargs.update({
'mask_prediction_fn': test_utils.MockMaskHead(num_classes=1).predict,
})
model = model_fn(
is_training=is_training,
anchor_generator=mock_anchor_generator,
box_predictor=mock_box_predictor,
box_coder=mock_box_coder,
feature_extractor=fake_feature_extractor,
encode_background_as_zeros=encode_background_as_zeros,
image_resizer_fn=image_resizer_fn,
non_max_suppression_fn=non_max_suppression_fn,
score_conversion_fn=score_conversion_fn,
classification_loss=classification_loss,
localization_loss=localization_loss,
classification_loss_weight=classification_loss_weight,
localization_loss_weight=localization_loss_weight,
normalize_loss_by_num_matches=normalize_loss_by_num_matches,
hard_example_miner=hard_example_miner,
target_assigner_instance=target_assigner_instance,
add_summaries=False,
normalize_loc_loss_by_codesize=normalize_loc_loss_by_codesize,
freeze_batchnorm=False,
inplace_batchnorm_update=False,
add_background_class=add_background_class,
random_example_sampler=random_example_sampler,
expected_loss_weights_fn=expected_loss_weights_fn,
**kwargs)
return model, num_classes, mock_anchor_generator.num_anchors(), code_size
def create_default_faster_rcnn_model_proto(self):
"""Creates a DetectionModel proto with FasterRCNN model fields populated."""
model_text_proto = """
faster_rcnn {
inplace_batchnorm_update: false
num_classes: 3
image_resizer {
keep_aspect_ratio_resizer {
min_dimension: 600
max_dimension: 1024
}
}
first_stage_anchor_generator {
grid_anchor_generator {
scales: [0.25, 0.5, 1.0, 2.0]
aspect_ratios: [0.5, 1.0, 2.0]
height_stride: 16
width_stride: 16
}
}
first_stage_box_predictor_conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
initial_crop_size: 14
maxpool_kernel_size: 2
maxpool_stride: 2
second_stage_box_predictor {
mask_rcnn_box_predictor {
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
fc_hyperparams {
op: FC
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
}
second_stage_post_processing {
batch_non_max_suppression {
score_threshold: 0.01
iou_threshold: 0.6
max_detections_per_class: 100
max_total_detections: 300
}
score_converter: SOFTMAX
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
(model_proto.faster_rcnn.feature_extractor.type
) = self.default_faster_rcnn_feature_extractor()
return model_proto
def test_create_faster_rcnn_resnet101_with_mask_prediction_enabled(self):
model_text_proto = """
faster_rcnn {
num_classes: 3
image_resizer {
keep_aspect_ratio_resizer {
min_dimension: 600
max_dimension: 1024
}
}
feature_extractor {
type: 'faster_rcnn_resnet101'
}
first_stage_anchor_generator {
grid_anchor_generator {
scales: [0.25, 0.5, 1.0, 2.0]
aspect_ratios: [0.5, 1.0, 2.0]
height_stride: 16
width_stride: 16
}
}
first_stage_box_predictor_conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
initial_crop_size: 14
maxpool_kernel_size: 2
maxpool_stride: 2
second_stage_box_predictor {
mask_rcnn_box_predictor {
fc_hyperparams {
op: FC
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
predict_instance_masks: true
}
}
second_stage_mask_prediction_loss_weight: 3.0
second_stage_post_processing {
batch_non_max_suppression {
score_threshold: 0.01
iou_threshold: 0.6
max_detections_per_class: 100
max_total_detections: 300
}
score_converter: SOFTMAX
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
model = model_builder.build(model_proto, is_training=True)
self.assertAlmostEqual(model._second_stage_mask_loss_weight, 3.0)
def test_create_ssd_mobilenet_v2_fpnlite_model_from_config(self):
model_text_proto = """
ssd {
freeze_batchnorm: true
inplace_batchnorm_update: true
feature_extractor {
type: 'ssd_mobilenet_v2_fpn'
use_depthwise: true
fpn {
min_level: 3
max_level: 7
additional_layer_depth: 128
}
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
box_coder {
faster_rcnn_box_coder {
}
}
matcher {
argmax_matcher {
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
ssd_anchor_generator {
aspect_ratios: 1.0
}
}
image_resizer {
fixed_shape_resizer {
height: 320
width: 320
}
}
box_predictor {
convolutional_box_predictor {
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
}
normalize_loc_loss_by_codesize: true
loss {
classification_loss {
weighted_softmax {
}
}
localization_loss {
weighted_smooth_l1 {
}
}
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
model = self.create_model(model_proto)
self.assertIsInstance(model, ssd_meta_arch.SSDMetaArch)
self.assertIsInstance(model._feature_extractor,
SSDMobileNetV2FpnFeatureExtractor)
self.assertTrue(model._normalize_loc_loss_by_codesize)
self.assertTrue(model._freeze_batchnorm)
self.assertTrue(model._inplace_batchnorm_update)
def test_create_center_net_model(self):
"""Test building a CenterNet model from proto txt."""
proto_txt = """
center_net {
num_classes: 10
feature_extractor {
type: "resnet_v2_101"
channel_stds: [4, 5, 6]
bgr_ordering: true
}
image_resizer {
keep_aspect_ratio_resizer {
min_dimension: 512
max_dimension: 512
pad_to_max_dimension: true
}
}
}
"""
# Set up the configuration proto.
config = text_format.Merge(proto_txt, model_pb2.DetectionModel())
config.center_net.object_center_params.CopyFrom(
self.get_fake_object_center_proto())
config.center_net.object_detection_task.CopyFrom(
self.get_fake_object_detection_proto())
config.center_net.keypoint_estimation_task.append(
self.get_fake_keypoint_proto())
config.center_net.keypoint_label_map_path = (
self.get_fake_label_map_file_path())
config.center_net.mask_estimation_task.CopyFrom(
self.get_fake_mask_proto())
config.center_net.densepose_estimation_task.CopyFrom(
self.get_fake_densepose_proto())
# Build the model from the configuration.
model = model_builder.build(config, is_training=True)
# Check object center related parameters.
self.assertEqual(model._num_classes, 10)
self.assertIsInstance(model._center_params.classification_loss,
losses.PenaltyReducedLogisticFocalLoss)
self.assertEqual(model._center_params.classification_loss._alpha, 3.0)
self.assertEqual(model._center_params.classification_loss._beta, 4.0)
self.assertAlmostEqual(model._center_params.min_box_overlap_iou, 0.2)
self.assertAlmostEqual(model._center_params.heatmap_bias_init,
3.14,
places=4)
self.assertEqual(model._center_params.max_box_predictions, 15)
# Check object detection related parameters.
self.assertAlmostEqual(model._od_params.offset_loss_weight, 0.1)
self.assertAlmostEqual(model._od_params.scale_loss_weight, 0.2)
self.assertAlmostEqual(model._od_params.task_loss_weight, 0.5)
self.assertIsInstance(model._od_params.localization_loss,
losses.L1LocalizationLoss)
# Check keypoint estimation related parameters.
kp_params = model._kp_params_dict['human_pose']
self.assertAlmostEqual(kp_params.task_loss_weight, 0.9)
self.assertAlmostEqual(kp_params.keypoint_regression_loss_weight, 1.0)
self.assertAlmostEqual(kp_params.keypoint_offset_loss_weight, 0.5)
self.assertAlmostEqual(kp_params.heatmap_bias_init, 2.14, places=4)
self.assertEqual(kp_params.classification_loss._alpha, 3.0)
self.assertEqual(kp_params.keypoint_indices, [0, 1, 2, 3])
self.assertEqual(kp_params.keypoint_labels,
['nose', 'left_shoulder', 'right_shoulder', 'hip'])
self.assertAllClose(kp_params.keypoint_std_dev, [0.3, 1.0, 1.0, 0.0])
self.assertEqual(kp_params.classification_loss._beta, 4.0)
self.assertIsInstance(kp_params.localization_loss,
losses.L1LocalizationLoss)
self.assertAlmostEqual(kp_params.keypoint_candidate_score_threshold,
0.3)
self.assertEqual(kp_params.num_candidates_per_keypoint, 12)
self.assertEqual(kp_params.peak_max_pool_kernel_size, 5)
self.assertAlmostEqual(kp_params.unmatched_keypoint_score, 0.05)
self.assertAlmostEqual(kp_params.box_scale, 1.7)
self.assertAlmostEqual(kp_params.candidate_search_scale, 0.2)
self.assertEqual(kp_params.candidate_ranking_mode,
'score_distance_ratio')
self.assertEqual(kp_params.offset_peak_radius, 3)
self.assertEqual(kp_params.per_keypoint_offset, True)
self.assertEqual(kp_params.predict_depth, True)
self.assertEqual(kp_params.per_keypoint_depth, True)
self.assertAlmostEqual(kp_params.keypoint_depth_loss_weight, 0.3)
# Check mask related parameters.
self.assertAlmostEqual(model._mask_params.task_loss_weight, 0.7)
self.assertIsInstance(model._mask_params.classification_loss,
losses.WeightedSoftmaxClassificationLoss)
self.assertEqual(model._mask_params.mask_height, 8)
self.assertEqual(model._mask_params.mask_width, 8)
self.assertAlmostEqual(model._mask_params.score_threshold, 0.7)
self.assertAlmostEqual(model._mask_params.heatmap_bias_init,
-2.0,
places=4)
# Check DensePose related parameters.
self.assertEqual(model._densepose_params.class_id, 0)
self.assertIsInstance(model._densepose_params.classification_loss,
losses.WeightedSoftmaxClassificationLoss)
self.assertIsInstance(model._densepose_params.localization_loss,
losses.L1LocalizationLoss)
self.assertAlmostEqual(model._densepose_params.part_loss_weight, 1.0)
self.assertAlmostEqual(model._densepose_params.coordinate_loss_weight,
2.0)
self.assertEqual(model._densepose_params.num_parts, 24)
self.assertAlmostEqual(model._densepose_params.task_loss_weight, 0.5)
self.assertTrue(model._densepose_params.upsample_to_input_res)
self.assertEqual(model._densepose_params.upsample_method, 'bilinear')
self.assertAlmostEqual(model._densepose_params.heatmap_bias_init,
-2.0,
places=4)
# Check feature extractor parameters.
self.assertIsInstance(
model._feature_extractor, center_net_resnet_feature_extractor.
CenterNetResnetFeatureExtractor)
self.assertAllClose(model._feature_extractor._channel_means, [0, 0, 0])
self.assertAllClose(model._feature_extractor._channel_stds, [4, 5, 6])
self.assertTrue(model._feature_extractor._bgr_ordering)
def _load_model_proto(filename):
model_proto = model_pb2.DetectionModel()
with open(filename, 'r') as fp:
text_format.Merge(fp.read(), model_proto)
return model_proto
def test_create_ssd_vgg16_model_from_config(self):
model_text_proto = """
ssd {
feature_extractor {
type: 'ssd_vgg_16'
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
box_coder {
faster_rcnn_box_coder {
}
}
matcher {
argmax_matcher {
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
ssd_anchor_generator {
aspect_ratios: 1.0
}
}
image_resizer {
fixed_shape_resizer {
height: 300
width: 300
}
}
box_predictor {
convolutional_box_predictor {
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
}
loss {
classification_loss {
weighted_softmax {
}
}
localization_loss {
weighted_smooth_l1 {
}
}
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
model = self.create_model(model_proto)
self.assertIsInstance(model, ssd_meta_arch.SSDMetaArch)
self.assertIsInstance(model._feature_extractor,
SSDVgg16FeatureExtractor)
image = tf.placeholder(tf.float32, shape=[32, 300, 300, 3])
print model._feature_extractor.extract_features(image)
import sys
sys.path.append('external')
from object_detection.core import target_assigner
from object_detection.builders import box_coder_builder, matcher_builder, box_predictor_builder
from object_detection.builders import anchor_generator_builder, losses_builder
from object_detection.builders import image_resizer_builder, post_processing_builder
from object_detection.builders import region_similarity_calculator_builder as sim_calc_builder
from object_detection.protos import model_pb2
from object_detection.core import box_list
from google.protobuf import text_format
import tensorflow as tf
# input
model = model_pb2.DetectionModel()
f = open('model.config', 'r')
text_format.Merge(f.read(), model)
f.close()
num_classes = 20
groundtruth_class = tf.get_variable('groundtruth_class', shape=[24, 5, 20])
groundtruth_box = tf.get_variable('groundtruth_box', shape=[24, 5, 4])
groundtruth_classes_with_background_list = [
tf.pad(one_hot_encoding, [[0, 0], [1, 0]], mode='CONSTANT')
for one_hot_encoding in tf.unstack(groundtruth_class)
]
groundtruth_boxlists = [
box_list.BoxList(boxes) for boxes in tf.unstack(groundtruth_box)
]
# construct models
def test_create_ssd_resnet_v1_fpn_model_from_config(self):
model_text_proto = """
ssd {
feature_extractor {
type: 'ssd_resnet50_v1_fpn'
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
batch_norm_trainable: true
}
box_coder {
faster_rcnn_box_coder {
}
}
matcher {
argmax_matcher {
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
multiscale_anchor_generator {
aspect_ratios: [1.0, 2.0, 0.5]
scales_per_octave: 2
}
}
image_resizer {
fixed_shape_resizer {
height: 320
width: 320
}
}
box_predictor {
weight_shared_convolutional_box_predictor {
depth: 32
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
num_layers_before_predictor: 1
}
}
loss {
classification_loss {
weighted_sigmoid_focal {
alpha: 0.25
gamma: 2.0
}
}
localization_loss {
weighted_smooth_l1 {
}
}
classification_weight: 1.0
localization_weight: 1.0
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
for extractor_type, extractor_class in SSD_RESNET_V1_FPN_FEAT_MAPS.items():
model_proto.ssd.feature_extractor.type = extractor_type
model = model_builder.build(model_proto, is_training=True)
self.assertIsInstance(model, ssd_meta_arch.SSDMetaArch)
self.assertIsInstance(model._feature_extractor, extractor_class)
def test_unknown_meta_architecture(self):
model_proto = model_pb2.DetectionModel()
with self.assertRaisesRegex(ValueError, 'Unknown meta architecture'):
model_builder.build(model_proto, is_training=True)
def test_create_embedded_ssd_mobilenet_v1_model_from_config(self):
model_text_proto = """
ssd {
feature_extractor {
type: 'embedded_ssd_mobilenet_v1'
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
batch_norm_trainable: true
}
box_coder {
faster_rcnn_box_coder {
}
}
matcher {
argmax_matcher {
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
ssd_anchor_generator {
aspect_ratios: 1.0
}
}
image_resizer {
fixed_shape_resizer {
height: 256
width: 256
}
}
box_predictor {
convolutional_box_predictor {
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
}
loss {
classification_loss {
weighted_softmax {
}
}
localization_loss {
weighted_smooth_l1 {
}
}
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
model = self.create_model(model_proto)
self.assertIsInstance(model, ssd_meta_arch.SSDMetaArch)
self.assertIsInstance(model._feature_extractor,
EmbeddedSSDMobileNetV1FeatureExtractor)
def create_default_ssd_model_proto(self):
"""Creates a DetectionModel proto with ssd model fields populated."""
model_text_proto = """
ssd {
feature_extractor {
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
box_coder {
faster_rcnn_box_coder {
}
}
matcher {
argmax_matcher {
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
ssd_anchor_generator {
aspect_ratios: 1.0
}
}
image_resizer {
fixed_shape_resizer {
height: 320
width: 320
}
}
box_predictor {
convolutional_box_predictor {
conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
}
loss {
classification_loss {
weighted_softmax {
}
}
localization_loss {
weighted_smooth_l1 {
}
}
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
model_proto.ssd.feature_extractor.type = (
self.default_ssd_feature_extractor())
return model_proto
def test_create_faster_rcnn_inception_resnet_v2_model_from_config(self):
model_text_proto = """
faster_rcnn {
num_classes: 3
image_resizer {
keep_aspect_ratio_resizer {
min_dimension: 600
max_dimension: 1024
}
}
feature_extractor {
type: 'faster_rcnn_inception_resnet_v2'
}
first_stage_anchor_generator {
grid_anchor_generator {
scales: [0.25, 0.5, 1.0, 2.0]
aspect_ratios: [0.5, 1.0, 2.0]
height_stride: 16
width_stride: 16
}
}
first_stage_box_predictor_conv_hyperparams {
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
initial_crop_size: 17
maxpool_kernel_size: 1
maxpool_stride: 1
second_stage_box_predictor {
mask_rcnn_box_predictor {
fc_hyperparams {
op: FC
regularizer {
l2_regularizer {
}
}
initializer {
truncated_normal_initializer {
}
}
}
}
}
second_stage_post_processing {
batch_non_max_suppression {
score_threshold: 0.01
iou_threshold: 0.6
max_detections_per_class: 100
max_total_detections: 300
}
score_converter: SOFTMAX
}
}"""
model_proto = model_pb2.DetectionModel()
text_format.Merge(model_text_proto, model_proto)
model = model_builder.build(model_proto, is_training=True)
self.assertIsInstance(model, faster_rcnn_meta_arch.FasterRCNNMetaArch)
self.assertIsInstance(
model._feature_extractor,
frcnn_inc_res.FasterRCNNInceptionResnetV2FeatureExtractor)
def testGetMaxNumContextFeaturesFromModelConfig(self):
model_config = model_pb2.DetectionModel()
model_config.faster_rcnn.context_config.max_num_context_features = 10
max_num_context_features = config_util.get_max_num_context_features(
model_config)
self.assertAllEqual(max_num_context_features, 10)
