o Si3 @sddlmZmZddlmZddlmZddlmZm Z m Z ddl Z ddl mZddlmZddlmZmZmZmZGd d d Z dd ed ed e edeefddZdS))Counter defaultdict)deepcopy)ceil)ListOptionalTupleN)CutSet)Cut)SecondsTimeSpanifnoneis_module_availablec@sFeZdZdZddefddZddd Zd eddfd d ZdddZ d S)CutSetStatisticsaJ Low-level utility for creating an overview human-readable description for a CutSet. It powers the :meth:`~lhotse.cut.set.CutSet.describe` utility. Unlike `describe`, it allows to gather stats for multiple cut sets in parallel and combine and display them later. Example workflow (typically, the loop in the second line would be parallelized):: >>> cut_sets = [CutSet(...), CutSet(...)] >>> stats = [CutSetStatistics().accumulate(cuts) for cuts in cut_sets] >>> total_stats = stats[0].combine(*stats[1:]) >>> total_stats.describe() FfullcCsltdstd||_tt|_tt|_|_g|_ gg|_ |_ |jr4tt |_ gg|_|_dSdS)NtabulatezSince Lhotse v1.11, this function requires the `tabulate` package to be installed. Please run 'pip install tabulate' to continue.)r ValueErrorrrintcountersr cut_custom sup_custom cut_durationsspeaking_time_durationsspeech_durationslistdurations_by_num_speakerssingle_durationsoverlapped_durations)selfrrG/home/ubuntu/.local/lib/python3.10/site-packages/lhotse/cut/describe.py__init__s  zCutSetStatistics.__init__otherreturnc Gst|}|D]T}|j|jksJddD]}t||D]}t|||t|||7<qqd|jr5dndD] }t||t||q8|jrZ|jD] }|j||j|qLq|S)zOCombinemultiple statistics into a new statistics object (does not modify self).zACannot combine statistics gathered with full=True and full=False.)rrr)rrr)rrr)rrgetattrextendr)rr"lhsrhsattrkrrr combine,s("   zCutSetStatistics.combinecutsc Csvdttdtfdd}|D]}|j|jt|dr-t|jdD] }|j |d7<q!|j dt |j 7<|j d t |j 7<|jD]$}|j|j|j d d7<t|jdD] }|j|d7<q`qH|j|t|dd |jr|j|t|ddd |j|t|d dd |jd|jdtd dD]}|j||t|||d qq |S)z1Gather statistics to display later for a cut set.segmentsr#cSstdd|DS)Ncss|]}|jVqdS)N)duration).0segmentrrr NszGCutSetStatistics.accumulate..total_duration_..)sum)r,rrr total_duration_Msz4CutSetStatistics.accumulate..total_duration_customr recordingsfeatures supervisions) min_speakers)r8 max_speakersN)rr floatrappendr-hasattrr r3rrr has_recording has_featurestrimmed_supervisionsrrr find_segments_with_speaker_countrrrrrange)rr+r2ckeysnum_spkrrr accumulateJsV    zCutSetStatistics.accumulateNc sddlm}dtdttttffdddtdtffdd }|j}t|}g}| d t |g| d ||g| d t |d g| d t |d g| dt |d g| dt|dd g| dt|d g| dt|dd g| dt|dd g| dt|dd g| dt|dd g| dt|d g|jD]\}}| |d|gqtdt||dd|jrtd|jD]\}}td |d!|d"q|jrtd#|jD]\}}td |d!|d"qt|j}t|j} ||} g} | d$||||d%d&g| d'|| | |d%d&g| d(|| | |d%d&g|jrt|j} t|j} | d)|| | |d%d*| |d%d+g| d,|| | |d%d*| |d%d+gtd-t|| dd|jsd.Sgd/g}|jD]&\}}t|}||}| |||||||d%|| d%gq| d0|||| d1d1gtd2t||d3dd4d.S)5z*Display accumulated statistics in the CLI.r)rsecondsr#cSs2t|d\}}t|d\}}t|t|t|fS)Ni<)divmodrr)rJhoursminutesrrr convert_sz+CutSetStatistics.describe..convert_cs(|\}}}|dd|dd|dS)N02d:r)rJhmrGrOrr time_as_str_sz/CutSetStatistics.describe..time_as_str_z Cuts count:zTotal duration (hh:mm:ss)meanz.1fstdminz25%z50%z75%Kz99%cz99.5%gX@z99.9%gX@maxz available:zCut statistics: fancy_grid)tablefmtzCUT custom fields:z- z (in z cuts)zSUPERVISION custom fields:zTotal speech durationz.2%z of recordingzTotal speaking time durationzTotal silence durationzSingle-speaker durationz (z of speech)zOverlapped speech durationzSpeech duration statistics:N)zNumber of speakerszDuration (hh:mm:ss)zSpeaking time (hh:mm:ss)z % of speechz% of speaking timeTotalz100.00%z1Speech duration statistics by number of speakers:firstrow)headersr^)rr rrstrrnparrayr1r>lenrVrWrX percentilemedianr\ritemstitleprintr most_commonrrrrrrr)rrrUr total_sum cut_statsrFval total_speechtotal_speaking_time total_silence speech_stats total_single total_overlap speaker_statsrH durations speaker_sum speaking_timerrTr describes      zCutSetStatistics.describe)F)r"rr#r)r#N) __name__ __module__ __qualname____doc__boolr!r*r rIryrrrr r s  ;rcutr8r9r#cCs|durtd}|dkr||ksJd|d|d|dkr,|tdkr,td|jgSt|jdkr@|dkr9gStd|jgSg}|d|jD]}||jdf||jd fqJ||jdf|jd d d d dt |D\}}|||d}d}d}g} |ddD]!\} } ||kr||kr| || f| dur|| rdnd7}| }qg} | D]%\} }| |krq| r| dd| kr| dd|f| d<q| | |fqdd| DS)a Given a Cut, find a list of intervals that contain the specified number of speakers. :param cuts: the Cut to search. :param min_speakers: the minimum number of speakers. :param max_speakers: the maximum number of speakers. :return: a list of TimeSpans. Ninfrz min_speakers=z and max_speakers=z are not valid.)NTFcSs|d|ddu|ddufS)Nrr4Tr)xrrr /sz2find_segments_with_speaker_count..)rFcSs g|] \}}|ddur|qS)r4Nr)r.itrrr 3s z4find_segments_with_speaker_count..r4rr;cSsg|] \}}t||qSr)r )r.startendrrr rMs) r=r r-rer7r>rrsort enumerate)rr8r9 timestampsr/ cut_start_idx cut_end_idx num_speakers seg_start intervals timestampis_startmerged_intervalsrrrrr rC sF    rC)rN) collectionsrrcopyrmathrtypingrrrnumpyrclhotser lhotse.cutr lhotse.utilsr r r rrrrCrrrr s(    ~