在 Tensorflow 中使用迁移学习是否需要预定义图像大小？

我打算使用预先训练的模型，例如fast_rcnn_resnet101_pets用于 Tensorflow 环境中的对象检测，如所述here https://tensorflow-object-detection-api-tutorial.readthedocs.io/en/latest/training.html

我为训练和测试集收集了几张图像。所有这些图像都有不同的尺寸。我必须将它们调整为通用尺寸吗？

faster_rcnn_resnet101_pets uses resnet with input size 224x224x3.

这是否意味着我必须在发送图像进行训练之前调整所有图像的大小？或者由 TF 自动处理。

python train.py --logtostderr --train_dir=training/ --pipeline_config_path=training/faster_rcnn_resnet101_pets.config

一般来说，使用相同大小的图像是一个好习惯吗？

不，您不需要自己将输入图像的大小调整为固定形状。Tensorflow 对象检测 api 有一个预处理步骤，可以调整所有输入图像的大小。以下是预处理步骤中定义的函数，其中有一个image_resizer_fn，它对应于一个名为image_resizer在配置内file https://github.com/tensorflow/models/blob/master/research/object_detection/inputs.py.

def transform_input_data(tensor_dict,
                     model_preprocess_fn,
                     image_resizer_fn,
                     num_classes,
                     data_augmentation_fn=None,
                     merge_multiple_boxes=False,
                     retain_original_image=False,
                     use_multiclass_scores=False,
                     use_bfloat16=False):


"""A single function that is responsible for all input data transformations.
  Data transformation functions are applied in the following order.
  1. If key fields.InputDataFields.image_additional_channels is present in
     tensor_dict, the additional channels will be merged into
     fields.InputDataFields.image.
  2. data_augmentation_fn (optional): applied on tensor_dict.
  3. model_preprocess_fn: applied only on image tensor in tensor_dict.
  4. image_resizer_fn: applied on original image and instance mask tensor in
     tensor_dict.
  5. one_hot_encoding: applied to classes tensor in tensor_dict.
  6. merge_multiple_boxes (optional): when groundtruth boxes are exactly the
     same they can be merged into a single box with an associated k-hot class
     label.

根据proto https://github.com/tensorflow/models/blob/master/research/object_detection/protos/image_resizer.proto文件中，您可以在 4 种不同的图像缩放器中进行选择，即

1. keep_aspect_ratio_resizer
2. 固定形状调整器
3. 身份调整器
4. 条件形状调整器

Here https://github.com/tensorflow/models/blob/master/research/object_detection/samples/configs/faster_rcnn_resnet101_pets.config是模型的示例配置文件faster_rcnn_resnet101_pets并且图像全部用 min_dimension=600 和 max_dimension=1024 重新整形

model {
  faster_rcnn {
    num_classes: 37
    image_resizer {
      keep_aspect_ratio_resizer {
        min_dimension: 600
        max_dimension: 1024
      }
    }
    feature_extractor {
      type: 'faster_rcnn_resnet101'
      first_stage_features_stride: 16
    }

事实上，调整大小的图像的形状对检测速度和准确度性能有很大影响。虽然对输入图像的大小没有具体要求，但最好所有最小尺寸的图像都大于合理值，以便卷积运算正常工作。