bundle.rd_plot_from_bundle

Author: Leonardo de Sousa Marques Affiliation: Embedded Computing Lab (ECL), Federal University of Santa Catarina (UFSC)

Description:

This module decodes and generates RD plots from a bundle of encoded Light Field files.

Observation:

It is crucial to have the repository cloned beforehand in the correct path so that the encoded file can be decoded. The bundle follows a hierarchical structure: bundle -> codec name -> lf name -> encoded files. RD report files follow the naming convention: {lf_name}_{codec_name}_rd_reports.json

get_codec_wrapper(codec_name, configuration, codec_configuration)[source]

Instantiate a codec wrapper, cloning and building the repository if needed.

Parameters:

  • codec_name (str) – Name of the codec

  • configuration (ConfigurationReader) – Configuration reader

  • codec_configuration (Dict) – Codec-specific configuration dictionary

Returns:

Instantiated codec wrapper

Return type:

CodecWrapper

get_codec_configuration(configuration, codec_name)[source]

Get codec configuration with inheritance resolution.

Parameters:

  • configuration (ConfigurationReader) – Configuration reader instance

  • codec_name (str) – Name of the codec

Returns:

Codec configuration dictionary

Raises:

ValueError – If codec not found in configuration

Return type:

Dict

compute_metrics(original_lf, decoded_lf, wrapper_instances)[source]

Compute quality metrics between original and decoded light fields.

Parameters:

  • original_lf (RAWLightFieldData) – Reference light field

  • decoded_lf (RAWLightFieldData) – Decoded light field

  • wrapper_instances (Dict) – Pre-instantiated quality tool wrappers

Returns:

Quality metric reports keyed by tool short-name

Return type:

Dict[str, Dict]

compute_metrics_from_encoded_and_decoded(codecs, configuration, encoded_files_dir, results_path)[source]

For each codec: decode all encoded files across all light fields, then compute quality metrics per light field and generate RD plots and JSON reports immediately after each light field is fully processed.

Parameters:

  • codecs (List[str]) – Codec names to process; empty list means all found

  • configuration (ConfigurationReader) – Configuration reader

  • encoded_files_dir (Path) – Root directory containing per-codec encoded files

  • results_path (Path) – Root directory where plots and reports will be saved

Returns:

Data organised as {codec: {lf_name: [(bpp, reports_by_wrapper)]}}

Return type:

Dict[str, Dict[str, List[Tuple[float, Dict[str, Dict]]]]]

load_bundle_data_from_rd_reports(rd_reports_dir, allowed_codecs, configuration)[source]

Build bundle data from pre-computed RD report JSON files.

Expects files named {lf_name}_{codec_name}_rd_reports.json containing a results dict keyed by target bitrate label, each entry having a rate field and per-metric dicts with a mean value.

The returned structure mirrors compute_metrics_from_encoded_and_decoded() so that generate_rd_plots_from_encoded() can consume both paths.

Parameters:

  • rd_reports_dir (Path) – Directory containing RD report JSON files

  • allowed_codecs (List[str]) – Codec names to include; empty list means all

  • configuration (ConfigurationReader) – Configuration reader

Returns:

Data organised as {codec: {lf_name: [(bpp, reports_by_wrapper)]}}

Raises:

ValueError – If rd_reports_dir does not exist

Return type:

Dict[str, Dict[str, List[Tuple[float, Dict[str, Dict]]]]]

generate_rd_plots_from_encoded(all_data, results_path, configuration, from_rd_reports=False)[source]

Generate R-D plots from bundle data for all configured light fields and metrics.

When from_rd_reports is True, metrics are taken from the report data itself and rd_plots config is used only for plot styling (a single pass).

Parameters:

  • all_data (Dict[str, Dict[str, List[Tuple[float, Dict[str, Dict]]]]]) – Bundle data organised as {codec: {lf_name: [(bpp, reports)]}}

  • results_path (Path) – Root directory where plots will be saved

  • configuration (ConfigurationReader) – Configuration reader

  • from_rd_reports (bool) – If True, derive metrics from report data instead of rd_plots config

Return type:

None

create_json_from_bundle(all_data, results_path, configuration, from_rd_reports=False)[source]

Generate JSON reports from bundle data containing rate-distortion information.

Creates detailed JSON files containing R-D values, metadata, and codec information organized by target bitrate and quality metric, mirroring the structure produced by the original create_json function.

Parameters:

  • all_data (Dict[str, Dict[str, List[Tuple[float, Dict[str, Dict]]]]]) – Bundle data organised as {codec: {lf_name: [(bpp, reports)]}}

  • results_path (Path) – Root directory where JSON reports will be saved

  • configuration (ConfigurationReader) – Configuration reader

  • from_rd_reports (bool) – If True, data comes from pre-computed RD reports

Return type:

None

main()[source]

Decode and generate R-D plots from an encoded light field bundle.

Reads configuration from command-line arguments, then either decodes encoded files and computes quality metrics, or loads pre-computed RD report JSON files, before generating rate-distortion plots.

Usage: python rd_plots_from_bundle.py <configuration.json>

Raises:

SystemExit – If configuration or required paths are invalid

Return type:

None