Module torpido.util.timestamp
This file will read the ranking for each section, parse it and create timestamps in seconds that can be used to clip the video using ffmpeg.
1. Reading the ranks and parsing them to timestamps
2. Padding the time ranks to generate the timestamps
3. Validating the timestamps
4. Returning a list of list containing (start -> end) time stamps
Expand source code
"""
This file will read the ranking for each section, parse it and
create timestamps in seconds that can be used to clip the video
using ffmpeg.
1. Reading the ranks and parsing them to timestamps
2. Padding the time ranks to generate the timestamps
3. Validating the timestamps
4. Returning a list of list containing (start -> end) time stamps
"""
import numpy as np
from torpido.config import *
from torpido.exceptions import RankingOfFeatureMissing
def addPadding(rankList: list, length):
"""
Function to add padding to the ranks if there length is lower than that of the
required length
Average of the rank is added as the padding data, mostly data is below 1
Parameters
----------
rankList : list
feature rank list
length : int
required length
"""
rankList.extend([np.average(rankList)] * int(length - len(rankList)))
def readTheRankings():
"""
Reads the ranking using the joblib files and calculate
the final sum ranks
Getting the ranking length for the ranks for the features from max
of all the ranks length
Padding the ranks, so that every feature has equal range
Returns
-------
list
list of the sum of all ranks
"""
if os.path.isfile(os.path.join(RANK_DIR, RANK_OUT_MOTION)) is False:
Log.e("Motion Ranking does not exists")
return None
if os.path.isfile(os.path.join(RANK_DIR, RANK_OUT_BLUR)) is False:
Log.e("Blur Ranking does not exists")
return None
if os.path.isfile(os.path.join(RANK_DIR, RANK_OUT_TEXT)) is False:
Log.e("Text Ranking does not exists")
return None
if os.path.isfile(os.path.join(RANK_DIR, RANK_OUT_AUDIO)) is False:
Log.e("Audio Ranking does not exists")
return None
# reading the saved ranking from the joblib files
motionFile = os.path.join(RANK_DIR, RANK_OUT_MOTION)
blurFile = os.path.join(RANK_DIR, RANK_OUT_BLUR)
textFile = os.path.join(RANK_DIR, RANK_OUT_TEXT)
audioFile = os.path.join(RANK_DIR, RANK_OUT_AUDIO)
# loading the read files
motionRank = load(motionFile)
blurRank = load(blurFile)
textRank = load(textFile)
audioRank = load(audioFile)
# max length for ranking
# NOTE: FFmpeg does not get bothered by greater values, not lower though
maxRank = int(max(len(motionRank), len(blurRank),
len(audioRank), len(textRank)))
# padding the ranks of each feature
addPadding(motionRank, maxRank)
addPadding(blurRank, maxRank)
addPadding(audioRank, maxRank)
addPadding(textRank, maxRank)
return [motionRank, blurRank,
textRank, audioRank]
def trimByRank(ranks):
"""
Parse the ranks to generate timestamps. Ranks are per sec so the start rank will be start
timestamp for trimming
Parameters
----------
ranks : iterable
ranks of the video
Returns
--------
timestamps : list
timestamps parsed from the ranks
"""
timestamps = []
start = None
end = None
prev_end = 0
for i in range(0, len(ranks)):
if ranks[i] > MIN_RANK_OUT_VIDEO:
if start is None:
start = i
prev_end += 1
else:
if end is None and start is not None:
end = start + prev_end - 1
if start == end:
end = None
if start is not None and end is not None:
start = start
end = end
timestamps.append([start, end])
end = None
start = None
prev_end = 0
if start is not None and end is None:
end = len(ranks)
timestamps.append([start, end])
return timestamps
def getTimestamps():
"""
Returns parsed timestamps from the ranking of all 4 processing
Returns
-------
list of list
timestamps list containing start and emd timestamps
"""
motion, blur, text, audio = readTheRankings()
ranks = [motion[i] + blur[i] + text[i] + audio[i] for i in range(len(motion))]
if ranks is not None:
timestamps = trimByRank(ranks)
else:
raise RankingOfFeatureMissing
finalTimestamp = []
# validating if there are 2 values in list
for clip in timestamps:
if len(clip) % 2 == 0:
finalTimestamp.append(clip)
return finalTimestamp
def getOutputVideoLength(timestamps: list):
"""
Calculates the output video length from the timestamps, each portion
length would be end - start
Parameters
----------
timestamps : list
timestamps for the video to edit
Returns
-------
int
final video length
References
----------
timestamps has list of start and end like :
[[start, end], [start, end]]
to get a length of each portion
end - start
to get final length
finalLength += end[i] - start[i] ; i: 0 -> len(timestamps)
"""
videoLength = 0
if len(timestamps) < 0:
return 0
for start, end in timestamps:
videoLength += abs(end - start)
return videoLengthFunctions
def addPadding(rankList: list, length)-
Function to add padding to the ranks if there length is lower than that of the required length Average of the rank is added as the padding data, mostly data is below 1
Parameters
rankList:list- feature rank list
length:int- required length
Expand source code
def addPadding(rankList: list, length): """ Function to add padding to the ranks if there length is lower than that of the required length Average of the rank is added as the padding data, mostly data is below 1 Parameters ---------- rankList : list feature rank list length : int required length """ rankList.extend([np.average(rankList)] * int(length - len(rankList))) def getOutputVideoLength(timestamps: list)-
Calculates the output video length from the timestamps, each portion length would be end - start
Parameters
timestamps:list- timestamps for the video to edit
Returns
int- final video length
References
timestamps has list of start and end like :
[[start, end], [start, end]]to get a length of each portion end - start
to get final length finalLength += end[i] - start[i] ; i: 0 -> len(timestamps)
Expand source code
def getOutputVideoLength(timestamps: list): """ Calculates the output video length from the timestamps, each portion length would be end - start Parameters ---------- timestamps : list timestamps for the video to edit Returns ------- int final video length References ---------- timestamps has list of start and end like : [[start, end], [start, end]] to get a length of each portion end - start to get final length finalLength += end[i] - start[i] ; i: 0 -> len(timestamps) """ videoLength = 0 if len(timestamps) < 0: return 0 for start, end in timestamps: videoLength += abs(end - start) return videoLength def getTimestamps()-
Returns parsed timestamps from the ranking of all 4 processing
Returns
listoflist- timestamps list containing start and emd timestamps
Expand source code
def getTimestamps(): """ Returns parsed timestamps from the ranking of all 4 processing Returns ------- list of list timestamps list containing start and emd timestamps """ motion, blur, text, audio = readTheRankings() ranks = [motion[i] + blur[i] + text[i] + audio[i] for i in range(len(motion))] if ranks is not None: timestamps = trimByRank(ranks) else: raise RankingOfFeatureMissing finalTimestamp = [] # validating if there are 2 values in list for clip in timestamps: if len(clip) % 2 == 0: finalTimestamp.append(clip) return finalTimestamp def readTheRankings()-
Reads the ranking using the joblib files and calculate the final sum ranks
Getting the ranking length for the ranks for the features from max of all the ranks length
Padding the ranks, so that every feature has equal range
Returns
list- list of the sum of all ranks
Expand source code
def readTheRankings(): """ Reads the ranking using the joblib files and calculate the final sum ranks Getting the ranking length for the ranks for the features from max of all the ranks length Padding the ranks, so that every feature has equal range Returns ------- list list of the sum of all ranks """ if os.path.isfile(os.path.join(RANK_DIR, RANK_OUT_MOTION)) is False: Log.e("Motion Ranking does not exists") return None if os.path.isfile(os.path.join(RANK_DIR, RANK_OUT_BLUR)) is False: Log.e("Blur Ranking does not exists") return None if os.path.isfile(os.path.join(RANK_DIR, RANK_OUT_TEXT)) is False: Log.e("Text Ranking does not exists") return None if os.path.isfile(os.path.join(RANK_DIR, RANK_OUT_AUDIO)) is False: Log.e("Audio Ranking does not exists") return None # reading the saved ranking from the joblib files motionFile = os.path.join(RANK_DIR, RANK_OUT_MOTION) blurFile = os.path.join(RANK_DIR, RANK_OUT_BLUR) textFile = os.path.join(RANK_DIR, RANK_OUT_TEXT) audioFile = os.path.join(RANK_DIR, RANK_OUT_AUDIO) # loading the read files motionRank = load(motionFile) blurRank = load(blurFile) textRank = load(textFile) audioRank = load(audioFile) # max length for ranking # NOTE: FFmpeg does not get bothered by greater values, not lower though maxRank = int(max(len(motionRank), len(blurRank), len(audioRank), len(textRank))) # padding the ranks of each feature addPadding(motionRank, maxRank) addPadding(blurRank, maxRank) addPadding(audioRank, maxRank) addPadding(textRank, maxRank) return [motionRank, blurRank, textRank, audioRank] def trimByRank(ranks)-
Parse the ranks to generate timestamps. Ranks are per sec so the start rank will be start timestamp for trimming
Parameters
ranks:iterable- ranks of the video
Returns
timestamps:list- timestamps parsed from the ranks
Expand source code
def trimByRank(ranks): """ Parse the ranks to generate timestamps. Ranks are per sec so the start rank will be start timestamp for trimming Parameters ---------- ranks : iterable ranks of the video Returns -------- timestamps : list timestamps parsed from the ranks """ timestamps = [] start = None end = None prev_end = 0 for i in range(0, len(ranks)): if ranks[i] > MIN_RANK_OUT_VIDEO: if start is None: start = i prev_end += 1 else: if end is None and start is not None: end = start + prev_end - 1 if start == end: end = None if start is not None and end is not None: start = start end = end timestamps.append([start, end]) end = None start = None prev_end = 0 if start is not None and end is None: end = len(ranks) timestamps.append([start, end]) return timestamps