github/tensorflow/tensorflow/contrib/slim/

TensorFlow-Slim

TF-Slim 是一個輕量級的庫，用來在TF中定義、訓練和評估復雜模型。tf-slim能夠自由混入原生TF和其它框架（如tf.contrib.learn中）。

用法

import tensorflow.contrib.slim as slim

為什么用TF-Slim?

TF-Slim中都有什么組成部分?

定義模型

變量

層

Scopes

實例: 實現VGG16

訓練模型

Training Tensorflow models requires a model, a loss function, the gradient computation and a training routine that iteratively computes the gradients of the model weights relative to the loss and updates the weights accordingly. TF-Slim provides both common loss functions and a set of helper functions that run the training and evaluation routines.

損失

The loss function defines a quantity that we want to minimize. For classification problems, this is typically the cross entropy between the true distribution and the predicted probability distribution across classes. For regression problems, this is often the sum-of-squares differences between the predicted and true values.

Certain models, such as multi-task learning models, require the use of multiple loss functions simultaneously. In other words, the loss function ultimately being minimized is the sum of various other loss functions. For example, consider a model that predicts both the type of scene in an image as well as the depth from the camera of each pixel. This model's loss function would be the sum of the classification loss and depth prediction loss.

TF-Slim provides an easy-to-use mechanism for defining and keeping track of loss functions via the losses module. Consider the simple case where we want to train the VGG network:

Training Loop

TF-Slim provides a simple but powerful set of tools for training models found in learning.py. These include a Train function that repeatedly measures the loss, computes gradients and saves the model to disk, as well as several convenience functions for manipulating gradients. For example, once we've specified the model, the loss function and the optimization scheme, we can call slim.learning.create_train_op and slim.learning.train to perform the optimization:

實例: 訓練VGG16模型

To illustrate this, let's examine the following sample of training the VGG network:

微調已存在的模型

 

Brief Recap on Restoring Variables from a Checkpoint

After a model has been trained, it can be restored using tf.train.Saver() which restores Variables from a given checkpoint. For many cases, tf.train.Saver() provides a simple mechanism to restore all or just a few variables.

 

Partially Restoring Models

It is often desirable to fine-tune a pre-trained model on an entirely new dataset or even a new task. In these situations, one can use TF-Slim's helper functions to select a subset of variables to restore:

 

Restoring models with different variable names

 

Fine-Tuning a Model on a different task

Consider the case where we have a pre-trained VGG16 model. The model was trained on the ImageNet dataset, which has 1000 classes. However, we would like to apply it to the Pascal VOC dataset which has only 20 classes. To do so, we can initialize our new model using the values of the pre-trained model excluding the final layer:

 

評估模型

Once we've trained a model (or even while the model is busy training) we'd like to see how well the model performs in practice. This is accomplished by picking a set of evaluation metrics, which will grade the model's performance, and the evaluation code which actually loads the data, performs inference, compares the results to the ground truth and records the evaluation scores. This step may be performed once or repeated periodically.

 

度量

我們定義一個度量來衡量訓練效果，這不是一個損失函數（損失被用來在訓練過程中進行優化的）。例如，我們訓練時最小化log損失，但是評估模型時我們也許會用 F1 score ,或者 Intersection Over Union score（這個值不可微，因此也不能用在損失函數上）。

TF-Slim提供了一組度量 operations。籠統地講，計算一個度量值可以被分為三部分：

1. 初始化：初始化用來計算度量的變量。
2. Aggregation: perform operations (sums, etc) used to compute the metrics.
3. Finalization: (optionally) perform any final operation to compute metric values. For example, computing means, mins, maxes, etc.

例如，計算mean_absolute_error，兩個變量 (count和total)被初始化為0。在 aggregation，我們得到一組predictions 和 labels，計算它們的絕對誤差並總計為total。我們每增加一組，count也隨之增加。最后，在 finalization階段 ，total除以count來獲得均值。

The following example demonstrates the API for declaring metrics. Because metrics are often evaluated on a test set which is different from the training set (upon which the loss is computed), we'll assume we're using test data:

images, labels = LoadTestData(...)
predictions = MyModel(images)

mae_value_op, mae_update_op = slim.metrics.streaming_mean_absolute_error(predictions, labels)
mre_value_op, mre_update_op = slim.metrics.streaming_mean_relative_error(predictions, labels)
pl_value_op, pl_update_op = slim.metrics.percentage_less(mean_relative_errors, 0.3)

就像例子描述的那樣，創建的metric返回兩個值： value_op 和 update_op。 value_op是一個 idempotent operation 返回metric的當前值。update_op 是一個 operation，它執行 aggregation步驟 並返回metric的值。

跟蹤value_op 和update_op 費時費力。為了解決這個問題，TF-Slim提供兩個方便的函數：

# 總計value和update ops 到兩個列表中:
value_ops, update_ops = slim.metrics.aggregate_metrics(
    slim.metrics.streaming_mean_absolute_error(predictions, labels),
    slim.metrics.streaming_mean_squared_error(predictions, labels))

# 總起value和update ops 到兩個字典中:
names_to_values, names_to_updates = slim.metrics.aggregate_metric_map({
    "eval/mean_absolute_error": slim.metrics.streaming_mean_absolute_error(predictions, labels),
    "eval/mean_squared_error": slim.metrics.streaming_mean_squared_error(predictions, labels),
})