otx.algorithms.anomaly.tools.sample

sample.py.

This is a sample python script showing how to train an end-to-end OTX Anomaly Classification Task.

Functions

main()	Run sample.py with given CLI arguments.
parse_args()	Parse CLI arguments.

Classes

OtxAnomalyTask(dataset_path, train_subset, ...)

OTX Anomaly Classification Task.

class otx.algorithms.anomaly.tools.sample.OtxAnomalyTask(dataset_path: str, train_subset: Dict[str, str], val_subset: Dict[str, str], test_subset: Dict[str, str], model_template_path: str, seed: int | None = None)

Bases: object

OTX Anomaly Classification Task.

Initialize OtxAnomalyTask.

Parameters:

• dataset_path (str) – Path to the MVTec dataset.

• train_subset (Dict[str, str]) – Dictionary containing path to train annotation file and path to dataset.

• val_subset (Dict[str, str]) – Dictionary containing path to validation annotation file and path to dataset.

• test_subset (Dict[str, str]) – Dictionary containing path to test annotation file and path to dataset.

• model_template_path (str) – Path to model template.

• seed (Optional[int]) – Setting seed to a value other than 0 also marks PytorchLightning trainer’s deterministic flag to True.

Example

>>> import os
>>> os.getcwd()
'~/otx/external/anomaly'

If MVTec dataset is placed under the above directory, then we could run,

>>> model_template_path = "./configs/classification/padim/template.yaml"
>>> dataset_path = "./datasets/MVTec"
>>> task = OtxAnomalyTask(
...     dataset_path=dataset_path,
...     train_subset={"ann_file": train.json, "data_root": dataset_path},
...     val_subset={"ann_file": val.json, "data_root": dataset_path},
...     test_subset={"ann_file": test.json, "data_root": dataset_path},
...     model_template_path=model_template_path
... )

>>> task.train()
Performance(score: 1.0, dashboard: (1 metric groups))

>>> task.export()
Performance(score: 0.9756097560975608, dashboard: (1 metric groups))

static clean_up() → None

Clean up the results directory used by anomalib.

create_task(task: str) → Any

Create base torch or openvino task.

Parameters:

task (str) – task type. Either base or openvino.

Returns:

Base Torch or OpenVINO Task Class.

Return type:

Any

Example

>>> self.create_task(task="base")
<anomaly_classification.torch_task.AnomalyClassificationTask>

create_task_environment() → TaskEnvironment

Create task environment.

static evaluate(task: IEvaluationTask, result_set: ResultSetEntity) → None

Evaluate the performance of the model.

Parameters:

• task (IEvaluationTask) – Task to evaluate the performance. Either torch or openvino.

• result_set (ResultSetEntity) – Results set containing the true and pred datasets.

export() → ModelEntity

Export the model via openvino.

export_nncf() → ModelEntity

Export NNCF model via openvino.

get_dataclass() → Type[AnomalyDetectionDataset] | Type[AnomalySegmentationDataset] | Type[AnomalyClassificationDataset]

Gets the dataloader based on the task type.

Raises:

ValueError – Validates task type.

Returns:

Dataloader

infer(task: IInferenceTask, output_model: ModelEntity) → ResultSetEntity

Get the predictions using the base Torch or OpenVINO tasks and models.

Parameters:

• task (IInferenceTask) – Task to infer. Either torch or openvino.

• output_model (ModelEntity) – Output model on which the weights are saved.

Returns:

Results set containing the true and pred datasets.

Return type:

ResultSetEntity

optimize() → None

Optimize the model via POT.

optimize_nncf() → None

Optimize the model via NNCF.

train() → ModelEntity

Train the base Torch model.

otx.algorithms.anomaly.tools.sample.main() → None

Run sample.py with given CLI arguments.

otx.algorithms.anomaly.tools.sample.parse_args() → Namespace

Parse CLI arguments.

Returns:

CLI arguments.

Return type:

(Namespace)