object - Tensorflow ssd-mobilenet-V2 training seems not progress well

Question

Now I'm training ssd_mobilenet_v2 net to detect car license plates from scratch. my loss graph after 300к steps looks like the huge saw teeth in log axis view with maximums on 5e+11. The lowest minimus was 130, then I've got a huge peak with 7e+11 again. Is It ok, or I've got some problem and my training process will never finish with good accuracy?

https://ibb.co/wCwTwLp the saw teeth

here is the pipeline.config file content

model {
  ssd {
    num_classes: 1
    image_resizer {
      fixed_shape_resizer {
        height: 300
        width: 300
      }
    }
    feature_extractor {
      type: "ssd_mobilenet_v2_keras"
      depth_multiplier: 1.0
      min_depth: 16
      conv_hyperparams {
        regularizer {
          l2_regularizer {
            weight: 3.9999998989515007e-05
          }
        }
        initializer {
          truncated_normal_initializer {
            mean: 0.0
            stddev: 0.029999999329447746
          }
        }
        activation: RELU_6
        batch_norm {
          decay: 0.9700000286102295
          center: true
          scale: true
          epsilon: 0.0010000000474974513
          train: true
        }
      }
      override_base_feature_extractor_hyperparams: true
    }
    box_coder {
      faster_rcnn_box_coder {
        y_scale: 10.0
        x_scale: 10.0
        height_scale: 5.0
        width_scale: 5.0
      }
    }
    matcher {
      argmax_matcher {
        matched_threshold: 0.5
        unmatched_threshold: 0.5
        ignore_thresholds: false
        negatives_lower_than_unmatched: true
        force_match_for_each_row: true
        use_matmul_gather: true
      }
    }
    similarity_calculator {
      iou_similarity {
      }
    }
    box_predictor {
      convolutional_box_predictor {
        conv_hyperparams {
          regularizer {
            l2_regularizer {
              weight: 3.9999998989515007e-05
            }
          }
          initializer {
            random_normal_initializer {
              mean: 0.0
              stddev: 0.009999999776482582
            }
          }
          activation: RELU_6
          batch_norm {
            decay: 0.9700000286102295
            center: true
            scale: true
            epsilon: 0.0010000000474974513
            train: true
          }
        }
        min_depth: 0
        max_depth: 0
        num_layers_before_predictor: 0
        use_dropout: false
        dropout_keep_probability: 0.800000011920929
        kernel_size: 1
        box_code_size: 4
        apply_sigmoid_to_scores: false
        class_prediction_bias_init: -4.599999904632568
      }
    }
    anchor_generator {
      ssd_anchor_generator {
        num_layers: 6
        min_scale: 0.1000000298023224
        max_scale: 0.549999988079071
        aspect_ratios: 1.0
        aspect_ratios: 1.0
        aspect_ratios: 0.5
        aspect_ratios: 1.0
        aspect_ratios: 0.33329999446868896
      }
    }
    post_processing {
      batch_non_max_suppression {
        score_threshold: 9.99999993922529e-09
        iou_threshold: 0.6000000238418579
        max_detections_per_class: 100
        max_total_detections: 100
        use_static_shapes: false
      }
      score_converter: SIGMOID
    }
    normalize_loss_by_num_matches: true
    loss {
      localization_loss {
        weighted_smooth_l1 {
          delta: 1.0
        }
      }
      classification_loss {
        weighted_sigmoid_focal {
          gamma: 2.0
          alpha: 0.75
        }
      }
      classification_weight: 1.0
      localization_weight: 1.0
    }
    encode_background_as_zeros: true
    normalize_loc_loss_by_codesize: true
    inplace_batchnorm_update: true
    freeze_batchnorm: false
  }
}
train_config {
  batch_size: 24
  data_augmentation_options {
    random_horizontal_flip {
    }
  }
  data_augmentation_options {
    ssd_random_crop {
    }
  }
  sync_replicas: true
  optimizer {
    momentum_optimizer {
      learning_rate {
        cosine_decay_learning_rate {
          learning_rate_base: 0.800000011920929
          total_steps: 1000000
          warmup_learning_rate: 0.13333000242710114
          warmup_steps: 2000
        }
      }
      momentum_optimizer_value: 0.8999999761581421
    }
    use_moving_average: false
  }
  #fine_tune_checkpoint: "modelsmy_ssd_mobilenet_V2_300ckpt-51"
  num_steps: 1000000
  startup_delay_steps: 0.0
  replicas_to_aggregate: 8
  max_number_of_boxes: 100
  unpad_groundtruth_tensors: false
  fine_tune_checkpoint_type: "detection"
  use_bfloat16: false
  fine_tune_checkpoint_version: V2
}
train_input_reader {
  label_map_path: "annotationslabelmap.pbtxt"
  tf_record_input_reader {
    input_path: "annotations/train.record"
  }
}
eval_config {
  metrics_set: "coco_detection_metrics"
  use_moving_averages: false
}
eval_input_reader {
  label_map_path: "annotationslabelmap.pbtxt"
  shuffle: false
  num_epochs: 1
  tf_record_input_reader {
    input_path: "annotations/test.record"
  }
}

question from:https://stackoverflow.com/questions/65647058/tensorflow-ssd-mobilenet-v2-training-seems-not-progress-well

Answer 1

Waitting for answers