Merge branch 'ci/split-output-files-b4-comparison-fixed' into 'main' (ce97dbf5) · Commits · IVAS Codec Public Collaboration / IVAS Codec

ci/basop-pages/create_report_pages.py

+1 −0

Original line number	Diff line number	Diff line
		@@ -37,6 +37,7 @@ Comparing:

		<br>
		<h2><a href="summary_{job_name}.html">Summary page</a></h2>
		<h2><a href="summary__split_{job_name}.html">Split comparison summary page</a></h2>
		<br>
		<br>

ci/basop-pages/create_summary_page.py

+31 −30

Original line number	Diff line number	Diff line
		@@ -4,11 +4,11 @@ from typing import List
		from create_report_pages import SUBPAGE_TMPL_CSS, FORMATS


		title = {
		TITLE_4_MEASURE = {
		"MLD": "Maximum MLD across channels",
		"DIFF": "Maximim absolute difference across channels",
		"SSNR": "Minimum SSNR across channels",
		"ODG": "Minimum PEAQ ODG across channels",
		"MAX_ABS_DIFF": "Maximum absolute difference across channels",
		"MIN_SSNR": "Minimum SSNR across channels",
		"MIN_ODG": "Minimum PEAQ ODG across channels",
		"DELTA_ODG": "PEAQ ODG using binauralized input and output",
		}

		@@ -16,9 +16,13 @@ SUMMARY_PAGE_TMPL_HTML = """

		<h1>Summary for job {job_name}, ID: {id_current}</h1>

		<hr>

		{images}

		"""
		IMAGE_HTML_TMPL = "<img src={image_dir}/histogram_{measure}_{format}.png>"
		SUBHEADING_HTML_TMP = "<h2>{subtitle}</h2>\n"


		def create_summary_page(
		@@ -26,53 +30,50 @@ def create_summary_page(
		id_current: int,
		job_name: str,
		measures: List[str],
		image_dir: str,
		):
		images = histogram_summary(job_name, measures)
		html = "\n<hr>\n".join(
		[
		SUBHEADING_HTML_TMP.format(subtitle=TITLE_4_MEASURE[m])
		+ " ".join(
		[
		IMAGE_HTML_TMPL.format(measure=m, format=f, image_dir=image_dir)
		for f in FORMATS
		]
		)
		for m in measures
		]
		)

		new_summary_page = SUBPAGE_TMPL_CSS + SUMMARY_PAGE_TMPL_HTML.format(
		id_current=id_current,
		job_name=job_name,
		images=images,
		images=html,
		)
		with open(html_out, "w") as f:
		f.write(new_summary_page)


		def histogram_summary(
		job_name: str,
		measures: List[str],
		):
		images = "<hr>"
		for m in measures:
		images += (
		f"<h2>{title[m]}</h2>\n"
		+ " ".join(
		[f"<img src=images_{job_name}/summary_{m}_{x}.png>" for x in FORMATS]
		)
		+ f'\n<br><a href="images_{job_name}/summary_{m}.csv">summary_{m}.csv</a><hr>\n\n'
		)
		return images


		if __name__ == "__main__":
		parser = argparse.ArgumentParser()
		parser.add_argument("html_out")
		parser.add_argument("id_current", type=int)
		parser.add_argument("job_name")
		parser.add_argument("image_dir")
		parser.add_argument(
		"--measures",
		nargs="+",
		help=f"List of measures to include in summary. Allowed values: {' '.join(title.keys())}",
		default=["MLD", "DIFF", "SSNR", "ODG"],
		help=f"List of measures to include in summary. Allowed values: {' '.join(TITLE_4_MEASURE.keys())}",
		# exclude DELTA_ODG here
		default=list(TITLE_4_MEASURE.keys())[:-1],
		)
		args = parser.parse_args()

		if not all([m in title for m in args.measures]):
		raise ValueError(f"Invalid list of measures: {args.measures}, expected one of {' '.join(title.keys())}")
		if not all([m in TITLE_4_MEASURE for m in args.measures]):
		raise ValueError(
		f"Invalid list of measures: {args.measures}, expected one of {' '.join(TITLE_4_MEASURE.keys())}"
		)

		create_summary_page(
		args.html_out,
		args.id_current,
		args.job_name,
		args.measures,
		args.html_out, args.id_current, args.job_name, args.measures, args.image_dir
		)

scripts/create_histogram_summary.py

deleted100644 → 0

+0 −156

Original line number	Diff line number	Diff line
		#!/usr/bin/env python3

		import argparse
		import math
		import numpy as np

		# These next three lines are added as a precaution in case the gitlab runner
		# needs DISPLAY to render the plots, even if they are written to file.
		import matplotlib

		matplotlib.use("Agg")
		import matplotlib.pyplot as plt
		import csv
		import os
		from parse_xml_report import IVAS_FORMATS, EVS_FORMATS, IVAS_CATEGORIES, EVS_CATEGORIES

		"""
		Parses a CSV report and creates a summary report.
		"""


		# Main routine
		if __name__ == "__main__":
		parser = argparse.ArgumentParser(
		description="Parses a CSV report and creates a summary report."
		)
		parser.add_argument(
		"csv_report",
		type=str,
		help="CSV report file of test cases, e.g. report.csv",
		)
		parser.add_argument(
		"csv_summary", type=str, help="Output CSV file, e.g. summary.csv"
		)
		parser.add_argument(
		"csv_image",
		type=str,
		nargs="?",
		help="Summary image file, e.g. summary.png",
		default=None,
		)
		parser.add_argument(
		"--measure",
		type=str,
		nargs=1,
		help="Measure, any of: MLD, DIFF, SSNR, ODG, default: MLD",
		default=["MLD"],
		)
		parser.add_argument(
		"--evs",
		action="store_true",
		help="Parse using EVS 26.444 formats",
		default=False,
		)
		parser.add_argument(
		"--diff",
		action="store_true",
		help="Use limits for diff scores",
		default=False,
		)
		args = parser.parse_args()
		csv_report = args.csv_report
		csv_summary = args.csv_summary
		csv_image = args.csv_image
		measure = args.measure[0]
		if args.evs:
		FORMATS = EVS_FORMATS
		CATEGORIES = EVS_CATEGORIES
		else:
		FORMATS = IVAS_FORMATS
		CATEGORIES = IVAS_CATEGORIES
		if args.diff:
		limits_per_measure = {
		"MLD": ("MLD", None),
		"DIFF": ("MAXIMUM ABS DIFF", None),
		"SSNR": ("MIN_SSNR", None),
		"ODG": ("MIN_ODG", None),
		"DELTA_ODG": ("DELTA_ODG", None),
		}
		else:
		limits_per_measure = {
		"MLD": ("MLD", [0, 1, 2, 3, 4, 5, 10, 20, math.inf]),
		"DIFF": ("MAXIMUM ABS DIFF", [0, 16, 256, 1024, 2048, 4096, 8192, 16384, 32769]),
		"SSNR": ("MIN_SSNR", [-math.inf, 0, 10, 20, 30, 40, 40, 50, 60, 100]),
		"ODG": ("MIN_ODG", [-5, -2, -1, -0.5, -0.4, -0.3, -0.2, -0.1, 0, 0.1, 0.5]),
		"DELTA_ODG": ("DELTA_ODG", [-5, -2, -1, -0.5, -0.4, -0.3, -0.2, -0.1, 0, 0.1, 0.5]),
		}
		(measure_label, limits) = limits_per_measure[measure]

		# Load CSV report
		results_sorted = {}
		with open(csv_report, "r") as fp:
		reader = csv.reader(fp, delimiter=";")
		header = next(reader)
		keys = header[1:]
		for row in reader:
		testcase = row[0]
		results_sorted[testcase] = {}
		for k, val in zip(keys, row[1:]):
		results_sorted[testcase][k] = val

		if limits is None:
		vals = [float(x) for x in [m[measure_label] for m in results_sorted.values() if m[measure_label] != "None" and m[measure_label] != ""]]
		start = min(vals)
		f = 10 ** (2 - int(np.floor(np.log10(abs(start)))) - 1)
		start = np.floor(start*f)/f
		step = (max(vals) - start)/10
		f = 10 ** (2 - int(np.floor(np.log10(abs(step)))) - 1)
		step = np.ceil(step*f)/f
		limits = np.arange(start, 10*step, step)

		# Output CSV file
		with open(csv_summary, "w") as fp:
		limits_labels = [f"{a:g}" for a in limits] + ["","None"] # Put None cases in separate bin
		headerline = f"Format;Category;" + ";".join(limits_labels) + "\n"
		fp.write(headerline)

		for fmt in FORMATS:
		fig, ax = plt.subplots()
		bottom = np.zeros(len(limits_labels))
		for cat in CATEGORIES:
		values = [
		x
		for x in [
		m[measure_label]
		for m in results_sorted.values()
		if m["Format"] == fmt and m["Category"] == cat
		]
		]
		# Create separate bin for None (errors)
		val = [float(x) for x in values if x != "None" and x != ""]
		none = [sum([1 for x in values if x == "None" or x == ""])]
		hist, _ = np.histogram(val, limits)
		data = np.array(list(hist) + [0] + none + [0])

		# CSV output
		line = f"{fmt};{cat};{'; '.join(map(str,data))}\n"
		fp.write(line)

		# Matplotlib histogram
		ax.bar(limits_labels, data, 1, align='edge', edgecolor='black', linewidth=0.5, label=cat, bottom=bottom)
		bottom += data

		# Histogram layout
		ax.set_title(fmt)
		ax.legend(loc="best")
		ax.set_xlabel(measure_label)
		if "DIFF" in measure_label:
		ax.set_xticks(range(len(limits_labels)), limits_labels, rotation=35)
		ax.set_ylabel("Number of test cases")

		fig.set_figheight(4)
		fig.set_figwidth(6)
		if csv_image:
		base, ext = os.path.splitext(csv_image)
		plt.savefig(f"{base}_{fmt}{ext}")

scripts/create_histograms.py

0 → 100644

+200 −0

Original line number	Diff line number	Diff line
		#!/usr/bin/env python3

		import argparse
		import math
		import pathlib
		import sys
		import pandas as pd
		import numpy as np
		from typing import List


		# hack for avoiding missing DISPLAY variable in headless CI runners
		import matplotlib

		matplotlib.use("Agg")
		import matplotlib.pyplot as plt


		BINS_FOR_MEASURES = {
		"MLD": [0, 1, 2, 3, 4, 5, 10, 20, math.inf],
		"MAX_ABS_DIFF": [0, 16, 256, 1024, 2048, 4096, 8192, 16384, 32769],
		"MIN_SSNR": [-math.inf, 0, 10, 20, 30, 40, 40, 50, 60, 100, math.inf],
		"MIN_ODG": [-5, -4, -3, -2, -1, -0.5, -0.4, -0.3, -0.2, -0.1, 0, 0.1, 0.5],
		"DELTA_ODG": [-5, -4, -3, -2, -1, -0.5, -0.4, -0.3, -0.2, -0.1, 0, 0.1, 0.5],
		}

		DEFAULT_MEASURES = ["MAX_ABS_DIFF", "MLD", "MIN_SSNR", "MIN_ODG"]

		### !!! Note: this is duplicated in tests/constatns.py. If you change this here, ALSO ADAPT IT THERE!!!
		### (importing from there failed for unknown reasons in some jobs on some runners and I don't have time to properly investigate this...)
		### below lines are the original solution, kept here for reference

		# HERE = pathlib.Path(__file__).parent
		# ROOT_DIR = HERE.parent
		# sys.path.append(str(ROOT_DIR))
		# from tests.constants import CAT_NORMAL, CAT_BITRATE_SWITCHING, CAT_DTX, CAT_JBM, CAT_PLC

		CAT_NORMAL = "normal operation"
		CAT_DTX = "DTX"
		CAT_PLC = "PLC"
		CAT_BITRATE_SWITCHING = "bitrate switching"
		CAT_JBM = "JBM"

		COLORS_FOR_CATEGORIES = {
		CAT_DTX: "tab:blue",
		CAT_PLC: "tab:orange",
		CAT_NORMAL: "tab:green",
		CAT_JBM: "tab:red",
		CAT_BITRATE_SWITCHING: "tab:purple",
		}


		def get_bins_for_diff(data: pd.Series):
		return np.round(np.linspace(data.min(), data.max(), num=10), decimals=2)


		def create_histograms(
		df: pd.DataFrame,
		measures: List[str],
		output_folder: pathlib.Path,
		display_only: bool,
		bins_for_measures=BINS_FOR_MEASURES,
		prefix="",
		write_out_histograms=False,
		):
		formats = df["format"].unique()
		categories = df["category"].unique()

		if not display_only or write_out_histograms:
		output_folder.mkdir(exist_ok=True, parents=True)

		for measure in measures:
		measure_in_df = prefix + measure
		bins = bins_for_measures.get(measure, get_bins_for_diff(df[measure_in_df]))
		x = [f"{x}" for x in bins] + ["", "ERROR"]

		df_hist = pd.DataFrame(columns=["format", "category"] + x)
		hist_row_count = 0

		for fmt in formats:
		fig, ax = plt.subplots()
		ax.xaxis.set_major_formatter("{x:.1f}")
		bottom = np.zeros(len(x))
		for cat in categories:
		data_mask = np.logical_and(df["format"] == fmt, df["category"] == cat)
		df_slice = df[data_mask]
		error_mask = df_slice["result"] == "ERROR"
		n_errors = np.sum(error_mask)
		df_slice = df_slice[np.logical_not(error_mask)]

		counts, _ = np.histogram(df_slice[measure_in_df], bins)

		data = np.concatenate([counts, [0], [n_errors], [0]])
		ax.bar(
		x,
		data,
		1,
		align="edge",
		edgecolor="black",
		linewidth=0.5,
		label=cat,
		bottom=bottom,
		color=COLORS_FOR_CATEGORIES[cat],
		)
		bottom += data

		hist_row = [fmt, cat] + list(counts) + [0] + [0, n_errors]
		df_hist.loc[hist_row_count] = hist_row
		hist_row_count += 1

		# Histogram layout
		ax.set_title(fmt)
		ax.legend(loc="best")
		ax.set_xlabel(measure)
		if "DIFF" in measure or len(bins_for_measures) == 0:
		ax.set_xticks(range(len(x)), x, rotation=35)
		else:
		ax.set_xticks(range(len(x)), x)
		ax.set_ylabel("Number of test cases")

		fig.set_figheight(4)
		fig.set_figwidth(6)
		plt.tight_layout()

		if not display_only:
		image_file = f"histogram_{measure}_{fmt}.png"
		image_path = output_folder.joinpath(image_file)
		plt.savefig(image_path)
		plt.close(fig)

		if write_out_histograms:
		df_hist.to_csv(
		output_folder.joinpath(f"histogram_{measure}.csv"), index=False
		)

		if display_only:
		plt.show()


		if __name__ == "__main__":
		parser = argparse.ArgumentParser(
		description="Parses a csv file generated by parse_xml_report and creates histograms for the given measures."
		)
		parser.add_argument(
		"csv_report",
		type=str,
		help="CSV report file as generated by parse_xml_report.py",
		)
		parser.add_argument(
		"output_folder", type=pathlib.Path, help="Output folder for writing the "
		)
		parser.add_argument(
		"--display-only",
		action="store_true",
		help="Do not write the output files, but display the graphs instead.",
		)
		parser.add_argument(
		"--no-bins",
		action="store_true",
		help="""Do not use the hardcoded bins for creating the spectrograms.
		Use this for visualising diff scores.""",
		)
		allowed_measures = " ".join(BINS_FOR_MEASURES.keys())
		parser.add_argument(
		"--measures",
		nargs="+",
		default=DEFAULT_MEASURES,
		help=f"Measures to plot from the csv file. One of {allowed_measures}",
		)
		parser.add_argument(
		"--prefix",
		default="",
		help="Common suffix to use when collecting measures from the input csv file",
		)
		parser.add_argument(
		"--write-out-histograms",
		action="store_true",
		help="Write out the histogram values to csv",
		)
		args = parser.parse_args()
		df = pd.read_csv(args.csv_report)

		# filter out missing format/category values
		mask_format_missing = df["format"].isna()
		mask_category_missing = df["category"].isna()
		df = df[~mask_format_missing \| ~mask_category_missing]

		bins_for_measures = BINS_FOR_MEASURES
		if args.no_bins:
		bins_for_measures = {}

		create_histograms(
		df,
		args.measures,
		args.output_folder,
		args.display_only,
		bins_for_measures,
		args.prefix,
		args.write_out_histograms,
		)

scripts/diff_report.py

+10 −6

Original line number	Diff line number	Diff line
		@@ -33,25 +33,29 @@ the United Nations Convention on Contracts on the International Sales of Goods.
		import pandas as pd
		import argparse
		import sys
		import os
		import pathlib

		COLUMNS_TO_COMPARE = [
		"MLD",
		"MAXIMUM ABS DIFF",
		"MAX_ABS_DIFF",
		"MIN_SSNR",
		"MIN_ODG",
		]


		def main(args):
		df_ref = pd.read_csv(args.csv_ref, sep=";")
		df_test = pd.read_csv(args.csv_test, sep=";")
		df_ref = pd.read_csv(args.csv_ref).sort_values(
		by=["testcase", "format", "category"]
		)
		df_test = pd.read_csv(args.csv_test).sort_values(
		by=["testcase", "format", "category"]
		)

		for col in COLUMNS_TO_COMPARE:
		df_ref[col] = df_test[col] - df_ref[col]
		df_ref.to_csv(args.csv_diff, index=False, sep=";")
		df_ref.to_csv(args.csv_diff, index=False)
		return 0


		if __name__ == "__main__":
		parser = argparse.ArgumentParser()
		parser.add_argument("csv_ref")