package fast_semantic_segmentation.protos

Get desktop application:
View/edit binary Protocol Buffers messages

Used in: Optimizer

• optional LearningRate learning_rate = 1

Used in: PreprocessingStep

• optional int32 min_dimension = 1

  Desired size of the smaller image dimension in pixels.

• optional int32 max_dimension = 2

  Desired size of the larger image dimension in pixels.

• optional ResizeType resize_method = 3

  Desired method when resizing image.

• optional bool pad_to_max_dimension = 4

  Whether to pad the image with zeros so the output spatial size is [max_dimension, max_dimension]. Note that the zeros are padded to the bottom and the right of the resized image.

Used in: Hyperparams

• optional float decay = 1

• optional bool center = 2

• optional bool scale = 3

• optional float epsilon = 4

• optional bool train = 5

Used in: Loss

• oneof loss_type

  • SoftmaxClassificationLoss softmax = 1

• optional CompressionStrategy compression_strategy = 1

Used in: CommpressionConfig

• oneof compression_strategy

  • FilterPruner filter_pruner = 1

Used in: LearningRate

• optional float learning_rate = 1

Used in: EvalInputType

• optional int32 height = 1

• optional int32 width = 2

Used in: PipelineConfig

• optional uint32 num_examples = 1

• optional uint32 ignore_label = 2

• optional EvalInputType eval_input_type = 3

Used in: EvalConfig

• oneof eval_input_type_oneof

  • CroppedEvalInput cropped_eval_input = 1

  • PaddedEvalInput padded_eval_input = 2

Used in: LearningRate

• optional float initial_learning_rate = 1

• optional uint32 decay_steps = 2

• optional float decay_factor = 3

• optional bool staircase = 4

Used in: PipelineConfig

• oneof model

  • PSPNet pspnet = 1

  • ICNet icnet = 2

Used in: CompressionStrategy

• optional NodeName input = 1

• optional NodeName output = 2

• optional string node_scope = 3

• optional float compression_factor = 4

• repeated NodeName skip_node = 5

• repeated SingleNodePruningSpec node = 6

Used in: ICNet, PSPNet

• optional Regularizer regularizer = 1

• optional Initializer initializer = 2

• optional BatchNorm batch_norm = 3

Used in: FastSegmentationModel

• optional int32 num_classes = 1

• optional PSPNetICNetFeatureExtractor feature_extractor = 2

• optional Hyperparams hyperparams = 3

• optional Loss loss = 4

  Multi-branch loss fields

• optional float main_branch_loss_weight = 5

• optional float second_branch_loss_weight = 6

• optional float first_branch_loss_weight = 7

• optional float filter_scale = 8

  The filter compression factor when applying filter/channel pruning

• repeated float pooling_factors = 9

  The PSP Module pooling factors. This is a list of reduction factors between (1.0, 0.0].

• optional bool pretrain_single_branch_mode = 10

  Used to pretrain a base PSPNet-like base model. This pretrained model should be use to initialize ICNet for training all branches.

• optional bool mobile_ops_only = 11

  prevents the use of tf.add_n for merging branches

Used in: Hyperparams

• oneof initializer_oneof

  • TruncatedNormalInitializer truncated_normal_initializer = 1

  • VarianceScalingInitializer variance_scaling_initializer = 2

Used in: PipelineConfig

• optional bool shuffle = 1

• optional uint32 queue_capacity = 2

• optional uint32 min_after_dequeue = 3

• optional uint32 num_epochs = 4

• optional uint32 num_readers = 5

• optional uint32 num_parallel_calls = 6

• optional uint32 shuffle_buffer = 7

• optional TFRecordInputReader tf_record_input_reader = 8

• optional uint32 num_examples = 9

Used in: Regularizer

• optional float weight = 1

Note that weight decay can be equivalent by setting the L2 lambda factor to the weight decay factor divited by 2.

Used in: Regularizer

• optional float weight = 1

Used in: AdamOptimizer, MomentumOptimizer

• oneof learning_rate

  • ConstantLearningRate constant_learning_rate = 1

  • ExponentialDecayLearningRate exponential_decay_learning_rate = 2

  • PolynomialDecayLearningRate polynomial_decay_learning_rate = 3

Used in: ICNet, PSPNet

• optional ClassificationLoss classification_loss = 1

• optional bool use_auxiliary_loss = 2

• optional uint32 ignore_label = 3

• optional bool upsample_logits = 4

Used in: Optimizer

• optional LearningRate learning_rate = 1

• optional float momentum_optimizer_value = 2

Used in: FilterPruner, SingleNodePruningSpec

• optional string name = 1

Used in: TrainConfig

• oneof optimizer

  • MomentumOptimizer momentum_optimizer = 1

  • AdamOptimizer adam_optimizer = 2

Used in: FastSegmentationModel

• optional int32 num_classes = 1

• optional PSPNetICNetFeatureExtractor feature_extractor = 2

• optional Hyperparams hyperparams = 3

• optional Loss loss = 4

• optional float filter_scale = 5

  The filter compression factor when applying filter/channel pruning

• repeated float pooling_factors = 6

  The PSP Module pooling factors. This is a list of reduction factors between (1.0, 0.0].

Used in: ICNet, PSPNet

• optional string type = 1

Used in: EvalInputType

• optional int32 height = 1

• optional int32 width = 2

• optional FastSegmentationModel model = 1

• optional TrainConfig train_config = 2

• optional InputReader train_input_reader = 3

• optional EvalConfig eval_config = 4

• optional InputReader eval_input_reader = 5

Used in: LearningRate

• optional float initial_learning_rate = 1

• optional uint32 decay_steps = 2

• optional float power = 3

Used in: TrainConfig

• oneof preprocessing_step

  • SetFixedImageSize set_fixed_image_size = 1

  • AspectRatioImageResize aspect_ratio_image_resize = 2

  • RandomImageScale random_image_scale = 3

  • RandomCropImage random_image_crop = 4

  • RandomHorizontalFlip random_horizontal_flip = 5

Used in: PreprocessingStep

• optional int32 crop_height = 1

• optional int32 crop_width = 2

• optional int32 images_channel_dim = 3

• optional int32 labels_channel_dim = 4

• optional int32 int_seed = 5

Used in: PreprocessingStep

(message has no fields)

Used in: PreprocessingStep

• optional float min_scale_ratio = 1

• optional float max_scale_ratio = 2

Used in: Hyperparams

• oneof regularizer_oneof

  • L1Regularizer l1_regularizer = 1

  • L2Regularizer l2_regularizer = 2

Used in: AspectRatioImageResize

• BILINEAR = 0

• NEAREST_NEIGHBOR = 1

• BICUBIC = 2

must be used for no resize to set image shape

Used in: PreprocessingStep

• optional int32 fixed_height = 1

• optional int32 fixed_width = 2

• optional int32 images_channel_dim = 3

• optional int32 labels_channel_dim = 4

Used in: FilterPruner

• optional NodeName target = 1

• optional NodeName source = 2

• repeated NodeName following = 3

• optional string node_scope = 4

Used in: ClassificationLoss

(message has no fields)

Used in: InputReader

• repeated string input_path = 1

Used in: PipelineConfig

• optional uint32 batch_size = 1

• repeated PreprocessingStep preprocessor_step = 2

• optional uint32 keep_checkpoint_every_n_hours = 3

• optional Optimizer optimizer = 4

• optional string fine_tune_checkpoint = 5

• optional string fine_tune_checkpoint_type = 6

  Type of checkpoint to restore variables from, e.g. 'classification' or 'segmentation'. Typically used to load feature extractor variables from trained models.

• optional bool freeze_fine_tune_backbone = 7

• optional uint32 num_steps = 8

  If 0, then train forever.

• optional int32 batch_queue_capacity = 9

• optional int32 num_batch_queue_threads = 10

• optional int32 prefetch_queue_capacity = 11

• optional bool add_regularization_loss = 12

• optional int32 quantize_with_delay = 13

• optional int32 num_queue_threads = 14

Used in: Initializer

• optional float mean = 1

• optional float stddev = 2

He initializatin

Used in: Initializer

• optional float factor = 1

• optional bool uniform = 2

• optional VarianceScalingInitializer.Mode mode = 3

Used in: VarianceScalingInitializer

• FAN_IN = 0

• FAN_OUT = 1

• FAN_AVG = 2