Adding more configurations

MusicObjectDetector/configurations/faster_rcnn_inception_v2_muscima_1.config

@@ -0,0 +1,143 @@
+# Faster R-CNN with Inception V2 feature extractor
+# Configured for MUSCIMA++ Dataset.
+# Loss oscillates and does not converge
+
+model {
+  faster_rcnn {
+    num_classes: 105
+    image_resizer {
+      keep_aspect_ratio_resizer {
+        min_dimension: 1800
+        max_dimension: 2700
+      }
+    }
+    feature_extractor {
+      type: 'faster_rcnn_inception_v2'
+      first_stage_features_stride: 8
+    }
+    first_stage_anchor_generator {
+      grid_anchor_generator {
+	  	width: 16
+		height: 16
+        scales: [0.25, 1.0, 4.0]
+        aspect_ratios: [0.125, 0.5, 1.0, 2.0, 8.0]
+        height_stride: 8
+        width_stride: 8
+      }
+    }
+    first_stage_atrous_rate: 2
+    first_stage_box_predictor_conv_hyperparams {
+      op: CONV
+      regularizer {
+        l2_regularizer {
+          weight: 0.0
+        }
+      }
+      initializer {
+        truncated_normal_initializer {
+          stddev: 0.01
+        }
+      }
+    }
+    first_stage_nms_score_threshold: 0.0
+    first_stage_nms_iou_threshold: 0.5
+    first_stage_max_proposals: 1000
+    first_stage_localization_loss_weight: 2.0
+    first_stage_objectness_loss_weight: 1.0
+    initial_crop_size: 17
+    maxpool_kernel_size: 1
+    maxpool_stride: 1
+    second_stage_box_predictor {
+      mask_rcnn_box_predictor {
+        use_dropout: false
+        dropout_keep_probability: 1.0
+        fc_hyperparams {
+          op: FC
+          regularizer {
+            l2_regularizer {
+              weight: 0.0
+            }
+          }
+          initializer {
+            variance_scaling_initializer {
+              factor: 1.0
+              uniform: true
+              mode: FAN_AVG
+            }
+          }
+        }
+      }
+    }
+    second_stage_post_processing {
+      batch_non_max_suppression {
+        score_threshold: 0.0
+        iou_threshold: 0.5
+        max_detections_per_class: 600
+        max_total_detections: 1000
+      }
+      score_converter: SOFTMAX
+    }
+    second_stage_localization_loss_weight: 2.0
+    second_stage_classification_loss_weight: 1.0
+  }
+}
+
+train_config: {
+  batch_size: 1
+  optimizer {
+    rms_prop_optimizer: {
+      learning_rate: {
+        exponential_decay_learning_rate {
+          initial_learning_rate: 0.003
+          decay_steps: 80000
+          decay_factor: 0.95
+        }
+      }
+      momentum_optimizer_value: 0.9
+      decay: 0.9
+      epsilon: 1.0
+    }
+  }
+  gradient_clipping_by_norm: 10.0
+  #fine_tune_checkpoint: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/faster_rcnn_inception_v2_coco_2018_01_28/model.ckpt"
+  #fine_tune_checkpoint_type:  "detection"
+  #from_detection_checkpoint: true
+  # Note: The below line limits the training process to 200K steps, which we
+  # empirically found to be sufficient enough to train the pets dataset. This
+  # effectively bypasses the learning rate schedule (the learning rate will
+  # never decay). Remove the below line to train indefinitely.
+  num_steps: 40000
+  data_augmentation_options {
+    random_horizontal_flip {
+    }
+  }
+
+  max_number_of_boxes: 1000
+}
+
+train_input_reader: {
+  tf_record_input_reader {
+    input_path: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/normalized/muscima/training.record"
+  }
+  label_map_path: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/normalized/muscima/mapping.txt"
+}
+
+eval_config: {
+  metrics_set: "pascal_voc_detection_metrics"
+  # metrics_set: "weighted_pascal_voc_detection_metrics"
+  #metrics_set: "coco_detection_metrics"
+  use_moving_averages: false
+  num_examples: 280
+  # Note: The below line limits the evaluation process to 10 evaluations.
+  # Remove the below line to evaluate indefinitely.
+  #max_evals: 1
+}
+
+eval_input_reader: {
+  tf_record_input_reader {
+    input_path: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/normalized/muscima/validation.record"
+  }
+  label_map_path: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/normalized/muscima/mapping.txt"
+  shuffle: false
+  num_readers: 1
+}

MusicObjectDetector/configurations/faster_rcnn_inception_v2_pretrained_muscima_1.config

@@ -0,0 +1,143 @@
+# Faster R-CNN with Inception V2 feature extractor
+# Configured for MUSCIMA++ Dataset.
+# Losses do not converge!
+
+model {
+  faster_rcnn {
+    num_classes: 105
+    image_resizer {
+      keep_aspect_ratio_resizer {
+        min_dimension: 1800
+        max_dimension: 2700
+      }
+    }
+    feature_extractor {
+      type: 'faster_rcnn_inception_v2'
+      first_stage_features_stride: 8
+    }
+    first_stage_anchor_generator {
+      grid_anchor_generator {
+	  	width: 16
+		height: 16
+        scales: [0.25, 1.0, 4.0]
+        aspect_ratios: [0.125, 0.5, 1.0, 2.0, 8.0]
+        height_stride: 8
+        width_stride: 8
+      }
+    }
+    first_stage_atrous_rate: 2
+    first_stage_box_predictor_conv_hyperparams {
+      op: CONV
+      regularizer {
+        l2_regularizer {
+          weight: 0.0
+        }
+      }
+      initializer {
+        truncated_normal_initializer {
+          stddev: 0.01
+        }
+      }
+    }
+    first_stage_nms_score_threshold: 0.0
+    first_stage_nms_iou_threshold: 0.5
+    first_stage_max_proposals: 1000
+    first_stage_localization_loss_weight: 2.0
+    first_stage_objectness_loss_weight: 1.0
+    initial_crop_size: 17
+    maxpool_kernel_size: 1
+    maxpool_stride: 1
+    second_stage_box_predictor {
+      mask_rcnn_box_predictor {
+        use_dropout: false
+        dropout_keep_probability: 1.0
+        fc_hyperparams {
+          op: FC
+          regularizer {
+            l2_regularizer {
+              weight: 0.0
+            }
+          }
+          initializer {
+            variance_scaling_initializer {
+              factor: 1.0
+              uniform: true
+              mode: FAN_AVG
+            }
+          }
+        }
+      }
+    }
+    second_stage_post_processing {
+      batch_non_max_suppression {
+        score_threshold: 0.0
+        iou_threshold: 0.5
+        max_detections_per_class: 600
+        max_total_detections: 1000
+      }
+      score_converter: SOFTMAX
+    }
+    second_stage_localization_loss_weight: 2.0
+    second_stage_classification_loss_weight: 1.0
+  }
+}
+
+train_config: {
+  batch_size: 1
+  optimizer {
+    rms_prop_optimizer: {
+      learning_rate: {
+        exponential_decay_learning_rate {
+          initial_learning_rate: 0.003
+          decay_steps: 80000
+          decay_factor: 0.95
+        }
+      }
+      momentum_optimizer_value: 0.9
+      decay: 0.9
+      epsilon: 1.0
+    }
+  }
+  gradient_clipping_by_norm: 10.0
+  fine_tune_checkpoint: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/faster_rcnn_inception_v2_coco_2018_01_28/model.ckpt"
+  fine_tune_checkpoint_type:  "detection"
+  from_detection_checkpoint: true
+  # Note: The below line limits the training process to 200K steps, which we
+  # empirically found to be sufficient enough to train the pets dataset. This
+  # effectively bypasses the learning rate schedule (the learning rate will
+  # never decay). Remove the below line to train indefinitely.
+  num_steps: 40000
+  data_augmentation_options {
+    random_horizontal_flip {
+    }
+  }
+
+  max_number_of_boxes: 1000
+}
+
+train_input_reader: {
+  tf_record_input_reader {
+    input_path: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/normalized/muscima/training.record"
+  }
+  label_map_path: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/normalized/muscima/mapping.txt"
+}
+
+eval_config: {
+  metrics_set: "pascal_voc_detection_metrics"
+  # metrics_set: "weighted_pascal_voc_detection_metrics"
+  #metrics_set: "coco_detection_metrics"
+  use_moving_averages: false
+  num_examples: 280
+  # Note: The below line limits the evaluation process to 10 evaluations.
+  # Remove the below line to evaluate indefinitely.
+  #max_evals: 1
+}
+
+eval_input_reader: {
+  tf_record_input_reader {
+    input_path: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/normalized/muscima/validation.record"
+  }
+  label_map_path: "C:/Users/Alex/Repositories/MusicObjectDetector-TF/MusicObjectDetector/data/normalized/muscima/mapping.txt"
+  shuffle: false
+  num_readers: 1
+}