Experiment module

Experiments are the main building blocks of the toolkit.

They are used to evaluate trackers on sequences in various ways.

class vot.experiment.Experiment(_identifier: str, _storage: Storage, **kwargs)

Experiment abstract base class.

Each experiment is responsible for running a tracker on a sequence and storing results into dedicated storage.

abstract execute(tracker: Tracker, sequence: Sequence, force: bool = False, callback: Callable = None)

Execute the experiment for a given tracker and sequence.

Parameters

  • tracker (Tracker) – Tracker to execute

  • sequence (Sequence) – Sequence to execute

  • force (bool) – Whether to force execution even if the results are already present

  • callback (Callable) – Callback to call after each frame

Returns

Results for the tracker and sequence

Return type

Results

property identifier: str

Identifier of the experiment.

Returns

Identifier of the experiment

Return type

str

log(identifier: str)

Get a log file for the experiment.

Parameters

identifier (str) – Identifier of the log

Returns

Path to the log file

Return type

str

results(tracker: Tracker, sequence: Sequence) → Results

Get results for a given tracker and sequence.

Parameters

  • tracker (Tracker) – Tracker to get results for

  • sequence (Sequence) – Sequence to get results for

Returns

Results for the tracker and sequence

Return type

Results

abstract scan(tracker: Tracker, sequence: Sequence)

Scan results for a given tracker and sequence.

Parameters

  • tracker (Tracker) – Tracker to scan results for

  • sequence (Sequence) – Sequence to scan results for

Returns

Results for the tracker and sequence

Return type

Results

property storage: Storage

Storage to use for storing results. Can be None if the experiment is not supposed to store results.

Returns

Storage to use for storing results

Return type

Storage

transform(sequences)

Transform a list of sequences using the experiment transformers.

Parameters

sequences (List[Sequence]) – List of sequences to transform

Returns

List of transformed sequences. The number of sequences may be larger than the input as some transformers may split sequences.

Return type

List[Sequence]

class vot.experiment.InjectConfig(*args, **kwargs)

Config proxy for parameter injection in experiments.

class vot.experiment.NoiseConfig(*args, **kwargs)

Config proxy for noise modifiers in experiments.

class vot.experiment.RealtimeConfig(*args, **kwargs)

Config proxy for real-time experiment.

vot.experiment.analysis_resolver(typename, context, **kwargs)

Resolve an analysis from a string. If the analysis is not registered, it is imported as a class and instantiated with the provided arguments.

Parameters

  • typename (str) – Name of the analysis

  • context (Attributee) – Context of the resolver

Returns

Resolved analysis

Return type

Analysis

vot.experiment.run_experiment(experiment: Experiment, tracker: Tracker, sequences: List[Sequence], force: bool = False, persist: bool = False)

A helper function that performs a given experiment with a given tracker on a list of sequences.

Parameters

  • experiment (Experiment) – The experiment object

  • tracker (Tracker) – The tracker object

  • sequences (List[Sequence]) – List of sequences.

  • force (bool, optional) – Ignore the cached results, rerun all the experiments. Defaults to False.

  • persist (bool, optional) – Continue runing even if exceptions were raised. Defaults to False.

Raises

TrackerException – If the experiment is interrupted

vot.experiment.transformer_resolver(typename, context, **kwargs)

Resolve a transformer from a string. If the transformer is not registered, it is imported as a class and instantiated with the provided arguments.

Parameters

  • typename (str) – Name of the transformer

  • context (Attributee) – Context of the resolver

Returns

Resolved transformer

Return type

Transformer