Question

我从一个大小为4864像素宽度，3648像素高度的JPG文件中总共获得5566条注释。我正在尝试使用预先训练的ssd_inception_v2_coco模型为我的数据集建立模型。

我的数据集包含作物田地上的谷物和非谷物的注释。注释（通过labelImg）很小，最小（非颗粒）注释的大小仅为2x3像素。但是，大多数注释的大小约为20x20像素。

在这里您可以看到我的配置文件：

# SSD with Inception v2 configuration for MSCOCO Dataset.
# Users should configure the fine_tune_checkpoint field in the train config as
# well as the label_map_path and input_path fields in the train_input_reader and
# eval_input_reader. Search for "PATH_TO_BE_CONFIGURED" to find the fields that
# should be configured.

model {
  ssd {
    num_classes: 1
    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
      }
    }
    similarity_calculator {
      iou_similarity {
      }
    }
    anchor_generator {
      ssd_anchor_generator {
        num_layers: 6
        min_scale: 0.2
        max_scale: 0.95
        aspect_ratios: 1.0
        aspect_ratios: 2.0
        aspect_ratios: 0.5
        aspect_ratios: 3.0
        aspect_ratios: 0.3333
        reduce_boxes_in_lowest_layer: true
      }
    }
    image_resizer {
      fixed_shape_resizer {
        height: 33
        width: 33
      }
    }
    box_predictor {
      convolutional_box_predictor {
        min_depth: 0
        max_depth: 0
        num_layers_before_predictor: 0
        use_dropout: false
        dropout_keep_probability: 0.8
        kernel_size: 3
        box_code_size: 4
        apply_sigmoid_to_scores: false
        conv_hyperparams {
          activation: RELU_6,
          regularizer {
            l2_regularizer {
              weight: 0.00004
            }
          }
          initializer {
            truncated_normal_initializer {
              stddev: 0.03
              mean: 0.0
            }
          }
        }
      }
    }
    feature_extractor {
      type: 'ssd_inception_v2'
      min_depth: 16
      depth_multiplier: 1.0
      conv_hyperparams {
        activation: RELU_6,
        regularizer {
          l2_regularizer {
            weight: 0.00004
          }
        }
        initializer {
          truncated_normal_initializer {
            stddev: 0.03
            mean: 0.0
          }
        }
        batch_norm {
          train: true,
          scale: true,
          center: true,
          decay: 0.9997,
          epsilon: 0.001,
        }
      }
      override_base_feature_extractor_hyperparams: true
    }
    loss {
      classification_loss {
        weighted_sigmoid {
        }
      }
      localization_loss {
        weighted_smooth_l1 {
        }
      }
      hard_example_miner {
        num_hard_examples: 3000
        iou_threshold: 0.99
        loss_type: CLASSIFICATION
        max_negatives_per_positive: 3
        min_negatives_per_image: 0
      }
      classification_weight: 1.0
      localization_weight: 1.0
    }
    normalize_loss_by_num_matches: true
    post_processing {
      batch_non_max_suppression {
        score_threshold: 1e-8
        iou_threshold: 0.6
        max_detections_per_class: 100
        max_total_detections: 100
      }
      score_converter: SIGMOID
    }
  }
}

train_config: {
  batch_size: 256
  optimizer {
    rms_prop_optimizer: {
      learning_rate: {
        exponential_decay_learning_rate {
          initial_learning_rate: 0.004
          decay_steps: 800720
          decay_factor: 0.95
        }
      }
      momentum_optimizer_value: 0.9
      decay: 0.9
      epsilon: 1.0
    }
  }
  fine_tune_checkpoint: "pre-trained-model/model.ckpt"
  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: 10000
  data_augmentation_options {
    random_horizontal_flip {
    }
  }
  data_augmentation_options {
    ssd_random_crop {
    }
  }
}

train_input_reader: {
  tf_record_input_reader {
    input_path: "annotations/train.record"
  }
  label_map_path: "annotations/label_map.pbtxt"
}

eval_config: {
  num_examples: 8000
  # Note: The below line limits the evaluation process to 10 evaluations.
  # Remove the below line to evaluate indefinitely.
  max_evals: 10
}

eval_input_reader: {
  tf_record_input_reader {
    input_path: "annotations/test.record"
  }
  label_map_path: "annotations/label_map.pbtxt"
  shuffle: false
  num_readers: 1
}

在这里您可以看到我的标签图：

item {
    id: 1
    name: 'grain'
}

item {
    id: 2
    name: 'nograin'
}

运行命令：python train.py --logtostderr --train_dir=training/ --pipeline_config_path=training/ssd_inception_v2_coco.config

后，您可以在此处看到Tensorflow的输出。

在我遇到OOM问题之前，这是Tensorflow的最后几行输出：

Instructions for updating:
Use standard file APIs to check for files with this prefix.
INFO:tensorflow:Restoring parameters from pre-trained-model/model.ckpt
INFO:tensorflow:Restoring parameters from pre-trained-model/model.ckpt
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Starting Session.
INFO:tensorflow:Starting Session.
INFO:tensorflow:Saving checkpoint to path training/model.ckpt
INFO:tensorflow:Saving checkpoint to path training/model.ckpt
INFO:tensorflow:Starting Queues.
INFO:tensorflow:Starting Queues.
INFO:tensorflow:global_step/sec: 0
INFO:tensorflow:global_step/sec: 0
Killed

这是我的内存使用情况（总共32GB RAM）：

我的问题是：如何处理此问题？可以通过更改配置文件来避免此问题吗？还是有一种降低模型复杂度的方法，以便不消耗所有内存或其他东西？

更新解决方案：：正如答案中所建议的，我将图像分为48个较小的部分，现在训练过程已启动并正在运行！

Answer 1

“来自单个JPG文件的5566注释，其尺寸为（4864像素宽度，3648像素高度）”-太大，您将无法使用该大小的图像做任何有意义的事情。请根据您要使用的网络的首选图像大小，将其分成较小的图像；如果您不能自行决定，则将其分成800x600。

如果图像分割重叠一点，例如在每种尺寸上都为100像素，这样可能会更好，因为它们会越过边框，因此不会丢失任何注释。

一旦图像被分割，使用任何最新的神经网络处理图像都不会出现任何问题。

处理Tensorflow中的内存不足问题

1 个答案: