page_type: reference description: A specification of the EfficientDet model.

tflite_model_maker.object_detector.EfficientDetSpec

<table class="tfo-notebook-buttons tfo-api nocontent" align="left"> <td> <a target="_blank" href="https://github.com/tensorflow/examples/blob/master/tensorflow_examples/lite/model_maker/core/task/model_spec/object_detector_spec.py#L109-L513">

View source on GitHub

</a> </td> </table>

A specification of the EfficientDet model.

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>tflite_model_maker.object_detector.EfficientDetSpec( model_name: str, uri: str, hparams: str = '', model_dir: Optional[str] = None, epochs: int = 50, batch_size: int = 64, steps_per_execution: int = 1, moving_average_decay: int = 0, var_freeze_expr: str = '(efficientnet|fpn_cells|resample_p6)', tflite_max_detections: int = 25, strategy: Optional[str] = None, tpu: Optional[str] = None, gcp_project: Optional[str] = None, tpu_zone: Optional[str] = None, use_xla: bool = False, profile: bool = False, debug: bool = False, tf_random_seed: int = 111111, verbose: int = 0 ) -> None </code></pre>   <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2"><h2 class="add-link">Args</h2></th></tr> <tr> <td> `model_name`<a id="model_name"></a> </td> <td> Model name. </td> </tr><tr> <td> `uri`<a id="uri"></a> </td> <td> TF-Hub path/url to EfficientDet module. </td> </tr><tr> <td> `hparams`<a id="hparams"></a> </td> <td> Hyperparameters used to overwrite default configuration. Can be

Dict, contains parameter names and values; 2) String, Comma separated k=v pairs of hyperparameters; 3) String, yaml filename which's a module containing attributes to use as hyperparameters.

</td> </tr><tr> <td> `model_dir`<a id="model_dir"></a> </td> <td> The location to save the model checkpoint files. </td> </tr><tr> <td> `epochs`<a id="epochs"></a> </td> <td> Default training epochs. </td> </tr><tr> <td> `batch_size`<a id="batch_size"></a> </td> <td> Training & Evaluation batch size. </td> </tr><tr> <td> `steps_per_execution`<a id="steps_per_execution"></a> </td> <td> Number of steps per training execution. </td> </tr><tr> <td> `moving_average_decay`<a id="moving_average_decay"></a> </td> <td> Float. The decay to use for maintaining moving averages of the trained parameters. </td> </tr><tr> <td> `var_freeze_expr`<a id="var_freeze_expr"></a> </td> <td> Expression to freeze variables. </td> </tr><tr> <td> `tflite_max_detections`<a id="tflite_max_detections"></a> </td> <td> The max number of output detections in the TFLite model. </td> </tr><tr> <td> `strategy`<a id="strategy"></a> </td> <td> A string specifying which distribution strategy to use. Accepted values are 'tpu', 'gpus', None. tpu' means to use TPUStrategy. 'gpus' mean to use MirroredStrategy for multi-gpus. If None, use TF default with OneDeviceStrategy. </td> </tr><tr> <td> `tpu`<a id="tpu"></a> </td> <td> The Cloud TPU to use for training. This should be either the name used when creating the Cloud TPU, or a grpc://ip.address.of.tpu:8470 url. </td> </tr><tr> <td> `gcp_project`<a id="gcp_project"></a> </td> <td> Project name for the Cloud TPU-enabled project. If not specified, we will attempt to automatically detect the GCE project from metadata. </td> </tr><tr> <td> `tpu_zone`<a id="tpu_zone"></a> </td> <td> GCE zone where the Cloud TPU is located in. If not specified, we will attempt to automatically detect the GCE project from metadata. </td> </tr><tr> <td> `use_xla`<a id="use_xla"></a> </td> <td> Use XLA even if strategy is not tpu. If strategy is tpu, always use XLA, and this flag has no effect. </td> </tr><tr> <td> `profile`<a id="profile"></a> </td> <td> Enable profile mode. </td> </tr><tr> <td> `debug`<a id="debug"></a> </td> <td> Enable debug mode. </td> </tr><tr> <td> `tf_random_seed`<a id="tf_random_seed"></a> </td> <td> Fixed random seed for deterministic execution across runs for debugging. </td> </tr><tr> <td> `verbose`<a id="verbose"></a> </td> <td> verbosity mode for <a href="https://www.tensorflow.org/api_docs/python/tf/keras/callbacks/ModelCheckpoint"><code>tf.keras.callbacks.ModelCheckpoint</code></a>, 0 or 1. </td> </tr> </table>

Methods

<h3 id="create_model"><code>create_model</code></h3>

<a target="_blank" class="external" href="https://github.com/tensorflow/examples/blob/master/tensorflow_examples/lite/model_maker/core/task/model_spec/object_detector_spec.py#L235-L238">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>create_model() -> tf.keras.Model </code></pre>

Creates the EfficientDet model.

<h3 id="evaluate"><code>evaluate</code></h3>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>evaluate( model: tf.keras.Model, dataset: tf.data.Dataset, steps: int, json_file: Optional[str] = None ) -> Dict[str, float] </code></pre>

Evaluate the EfficientDet keras model.

<table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Args</th></tr> <tr> <td> `model` </td> <td> The keras model to be evaluated. </td> </tr><tr> <td> `dataset` </td> <td> tf.data.Dataset used for evaluation. </td> </tr><tr> <td> `steps` </td> <td> Number of steps to evaluate the model. </td> </tr><tr> <td> `json_file` </td> <td> JSON with COCO data format containing golden bounding boxes. Used for validation. If None, use the ground truth from the dataloader. Refer to <a href="https://towardsdatascience.com/coco-data-format-for-object-detection-a4c5eaf518c5">https://towardsdatascience.com/coco-data-format-for-object-detection-a4c5eaf518c5</a> for the description of COCO data format. </td> </tr> </table>  <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Returns</th></tr> <tr class="alt"> <td colspan="2"> A dict contains AP metrics. </td> </tr> </table> <h3 id="evaluate_tflite"><code>evaluate_tflite</code></h3>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>evaluate_tflite( tflite_filepath: str, dataset: tf.data.Dataset, steps: int, json_file: Optional[str] = None ) -> Dict[str, float] </code></pre>

Evaluate the EfficientDet TFLite model.

<table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Args</th></tr> <tr> <td> `tflite_filepath` </td> <td> File path to the TFLite model. </td> </tr><tr> <td> `dataset` </td> <td> tf.data.Dataset used for evaluation. </td> </tr><tr> <td> `steps` </td> <td> Number of steps to evaluate the model. </td> </tr><tr> <td> `json_file` </td> <td> JSON with COCO data format containing golden bounding boxes. Used for validation. If None, use the ground truth from the dataloader. Refer to <a href="https://towardsdatascience.com/coco-data-format-for-object-detection-a4c5eaf518c5">https://towardsdatascience.com/coco-data-format-for-object-detection-a4c5eaf518c5</a> for the description of COCO data format. </td> </tr> </table>  <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Returns</th></tr> <tr class="alt"> <td colspan="2"> A dict contains AP metrics. </td> </tr> </table> <h3 id="export_saved_model"><code>export_saved_model</code></h3>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>export_saved_model( model: tf.keras.Model, saved_model_dir: str, batch_size: Optional[int] = None, pre_mode: Optional[str] = 'infer', post_mode: Optional[str] = 'global' ) -> None </code></pre>

Saves the model to Tensorflow SavedModel.

<table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Args</th></tr> <tr> <td> `model` </td> <td> The EfficientDetNet model used for training which doesn't have pre and post processing. </td> </tr><tr> <td> `saved_model_dir` </td> <td> Folder path for saved model. </td> </tr><tr> <td> `batch_size` </td> <td> Batch size to be saved in saved_model. </td> </tr><tr> <td> `pre_mode` </td> <td> Pre-processing Mode in ExportModel, must be {None, 'infer'}. </td> </tr><tr> <td> `post_mode` </td> <td> Post-processing Mode in ExportModel, must be {None, 'global', 'per_class', 'tflite'}. </td> </tr> </table> <h3 id="export_tflite"><code>export_tflite</code></h3>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>export_tflite( model: tf.keras.Model, tflite_filepath: str, quantization_config: Optional[<a href="../../tflite_model_maker/config/QuantizationConfig"><code>tflite_model_maker.config.QuantizationConfig</code></a>] = None ) -> None </code></pre>

Converts the retrained model to tflite format and saves it.

The exported TFLite model has the following inputs & outputs: One input: image: a float32 tensor of shape[1, height, width, 3] containing the normalized input image. self.config.image_size is [height, width].

<table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Four Outputs</th></tr> <tr> <td> `detection_boxes` </td> <td> a float32 tensor of shape [1, num_boxes, 4] with box locations. </td> </tr><tr> <td> `detection_classes` </td> <td> a float32 tensor of shape [1, num_boxes] with class indices. </td> </tr><tr> <td> `detection_scores` </td> <td> a float32 tensor of shape [1, num_boxes] with class scores. </td> </tr><tr> <td> `num_boxes` </td> <td> a float32 tensor of size 1 containing the number of detected boxes. </td> </tr> </table>  <table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2">Args</th></tr> <tr> <td> `model` </td> <td> The EfficientDetNet model used for training which doesn't have pre and post processing. </td> </tr><tr> <td> `tflite_filepath` </td> <td> File path to save tflite model. </td> </tr><tr> <td> `quantization_config` </td> <td> Configuration for post-training quantization. </td> </tr> </table> <h3 id="get_default_quantization_config"><code>get_default_quantization_config</code></h3>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>get_default_quantization_config( representative_data: <a href="../../tflite_model_maker/object_detector/DataLoader"><code>tflite_model_maker.object_detector.DataLoader</code></a> ) -> <a href="../../tflite_model_maker/config/QuantizationConfig"><code>tflite_model_maker.config.QuantizationConfig</code></a> </code></pre>

Gets the default quantization configuration.

<h3 id="train"><code>train</code></h3>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link"> <code>train( model: tf.keras.Model, train_dataset: tf.data.Dataset, steps_per_epoch: int, val_dataset: Optional[tf.data.Dataset], validation_steps: int, epochs: Optional[int] = None, batch_size: Optional[int] = None, val_json_file: Optional[str] = None ) -> tf.keras.Model </code></pre>

Run EfficientDet training.

<table class="responsive fixed orange"> <colgroup><col width="214px"><col></colgroup> <tr><th colspan="2"><h2 class="add-link">Class Variables</h2></th></tr> <tr> <td> compat_tf_versions<a id="compat_tf_versions"></a> </td> <td> `[2]` </td> </tr> </table>